Motor effects of cueing in Parkinson’s rehabilitation

Universidade do Minho Escola de Engenharia Maria Miguel Sá Marques Pimenta Motor effects of Cueing in Parkinson’s rehabilitation October 2024 Motor effects of Cueing in Parkinson’ s rehabilitation Maria Miguel Sá Marques Pimenta UMinho | 2024 Maria Miguel Sá Marques Pimenta Motor effects of Cueing in Parkinson’s Rehabilitation October 2024 Master’s Dissertation Master Degree in Biomedical Engineering Medical Electronics Branch Dissertation supervised by Professor Doctor Cristina Peixoto dos Santos i DIREITOS DE AUTOR E CONDIÇÕES DE UTILIZAÇÃO DO TRABALHO POR TERCEIROS Este é um trabalho académico que pode ser utilizado por terceiros desde que respeitadas as regras e boas práticas internacionalmente aceites, no que concerne aos direitos de autor e direitos conexos. Assim, o presente trabalho pode ser utilizado nos termos previstos na licença abaixo indicada. Caso o utilizador necessite de permissão para poder fazer um uso do trabalho em condições não previstas no licenciamento indicado, deverá contactar o autor, através do RepositóriUM da Universidade do Minho. Licença concedida aos utilizadores deste trabalho Atribuição CC BY https://creativecommons.org/licenses/by/4.0/ ii STATEMENT OF INTEGRITY I hereby declare having conducted this academic work with integrity. I confirm that I have not used plagiarism or any form of undue use of information or falsification of results along the process leading to its elaboration. I further declare that I have fully acknowledged the Code of Ethical Conduct of the University of Minho. iii AGRADECIMENTOS Antes de mais, peço desculpa por destoar esta secção das restantes colocando-a em português, mas realmente só faria sentido se assim o fosse. Este ano foi um ano de muito crescimento pessoal. Escrever uma dissertação sem dúvida foi uma tarefa que posso descrever como complicada, não pelo projeto em si porque como se costuma dizer “quem corre por gosto não cansa”, mas por exigir momentos mais solitários que não combinam tanto comigo. No entanto, foi bastante menos complicada por ter tido a sorte de ser acompanhada pelas pessoas certas. E como este trabalho não teria sido possível realizar sem a ajuda direta ou indireta dessas pessoas, merecem também estar aqui mencionadas. Primeiramente, gostaria de agradecer à minha orientadora, a professora Cristina, por todas as oportunidades, por ter acreditado em mim mesmo quando eu própria poderia achar que não seria capaz, e por ter me permitido trabalhar diretamente com pessoas excecionais, de onde saliento a Cristiana. A Cristiana foi sem dúvida essencial para todo o decorrer deste projeto, e um parágrafo ou uma página nunca serão agradecimento suficiente. Obrigada por todos os ensinamentos, por todo o tempo investido, e principalmente pela paciência para me aturar sempre. De seguida, quero agradecer aos meus pais, por apoiarem sempre todas as minhas decisões e me permitirem realizar as mesmas, e ainda à Ju, ao Luís e ao Scott, por serem o meu porto de abrigo constante. Agradeço ainda à restante família e amigos por estarem sempre presentes e disponíveis, inclusive para ser cobaias quando necessário. E como a tese não teria sido feita sem ter passado pelos restantes anos do curso, gostaria de agradecer também a todas as pessoas que o marcaram. Antes de mais, agradecer ao peste por todas as batalhas conquistadas juntos, em especial ao Kim, o meu parceiro maravilha na luta que foi o 1.º ano de mestrado. Às meninas, Demi, Pebbles, Red, Mary e Apita, agradeço por estarem sempre lá mesmo não precisando de estar, por me fazerem ver o copo meio cheio quando estou a olhar para o vazio, e por me fazerem entender o verdadeiro significado de amizade todos os dias. Resta-me também agradecer à Bia, por ser o sinónimo literal de casa, por todas as boleias, e por celebrar sempre todas as vitórias comigo por mais pequeninas que sejam. Todos vocês são especiais, e levar-vos-ei comigo sempre. Para terminar, gostaria de agradecer imensamente a todos os pacientes de Parkinson com os quais tive a oportunidade de trabalhar, pela disponibilidade, interesse e por tornarem este projeto realmente gratificante. A todos um enorme obrigada, e espero que este trabalho vos deixe tão felizes quanto me deixou a mim. iv RESUMO A doença de Parkinson (PD), é um distúrbio neurodegenerativo que afeta principalmente neurónios produtores de dopamina, traduzindo-se em sintomas motores incapacitantes como perturbações da marcha e freezing of gait (FoG) [1], [2]. O Parkinson impacta a autonomia dos pacientes ao realizar tarefas motoras diárias, diminuindo assim a sua qualidade de vida [2], [3]. Estratégias de aplicação de pistas sensoriais em dispositivos vestivéis, surgem como possibilidades promissoras para melhorar a qualidade de vida ( QoL ) [3]–[5]. Personalizadas para as necessidades dos pacientes de PD, permitem estimular diferentes caminhos neurológicos para gerar resposta motora [6]. As pistas podem assim ajudar os pacientes a restabelecer controlo motor e melhorar a mobilidade dos mesmos, surgindo como uma abordagem holística para os revigorar e restaurar a sua independência [5], [7]. Este projeto pretende avaliar o impacto das pistas sensoriais, focando-se no treino contínuo com a +sensBand em pacientes de PD , explorando o seu potencial para melhorar o controlo motor e reduzir episódios de FoG . A +sensBand é um dispositivo de biofeedback vestível, na forma de uma cinta, produzido pelo BiRDLaB , desenhado para potencializar reabilitação, monitorização motora e ainda administração de pistas táteis ou visuais, personalizada em pacientes de PD . O estudo tem como objetivo compreender os efeitos motores do uso de diferentes estratégias da +sensBand , implementando novas focadas em sintomas mais avançados como o FoG , e testando-as a longo prazo sobre os pacientes, através da estruturação e condução de um protocolo para um estudo longitudinal clínico, e consequente análise dos dados adquiridos. Os estudos de validação realizados demonstraram que as pistas vibrotáteis tanto em open como closed loop , melhoraram eficazmente o desempenho da marcha, ao abordar os aspetos hipocinéticos e bradicinéticos em pacientes de PD . A combinação de pistas visuais e somatossensoriais, de ambos os métodos, numa nova estratégia focada na prevenção de FoG , foi validada com elevada aceitação em participantes saudáveis, tendo comprovado o potencial promissor para a reabilitação motora em diferentes cenários da vida diária. Estes resultados sustentaram a sua aplicação a longo prazo com os pacientes, tendo a sua eficácia dependido de um treino consistente. Adicionalmente, o treino contínuo com esta estratégia reduziu as ocorrências de FoG , melhorou o desempenho motor e a QoL dos pacientes de PD , destacando o impacto positivo desta estratégia no quotidiano destes pacientes. Palavras-chave: Doença de Parkinson; estudo longitudinal; Freezing of Gait ; pistas sensoriais; Qualidade de vida v ABSTRACT Parkinson's Disease (PD) is a neurodegenerative disorder that affects dopaminergic neurons, leading to debilitating motor symptoms like gait disturbances and Freezing of Gait (FoG) [1], [2]. PD impacts patient's autonomy performing daily tasks, diminishing their overall quality of life (QoL) [2], [3]. Cueing strategies implemented on wearable devices, emerge as promising possibilities for enhancing QoL [3]–[5]. Tailored to the needs of PD patients, allow to explore different neurological pathways to generate motor response [6]. Cueing strategies can help PD patients regain motor control and improve mobility, offering a holistic approach that empowers patients and restores independence [5], [7]. This project aims to assess the impact of cueing strategies, focusing on the use of +sensBand by PD patients during several sessions, exploring its potential in improving motor control and reducing FoG episodes. The +sensBand is a wearable cueing system, an instrumented waistband from BiRDLab designed to enhance rehabilitation, motor assistence, and personalized clinical management on PD patients. The study aims to understand the motor effects of different +sensBand strategies, implement new strategies focused on targeting more advanced symptoms such as FoG, and testing the long-term rehabilitative potential of the system in PD patients by designing and conducting a protocol for a longitudinal clinical study and consequent analysis of collected data. The validation studies conducted demonstrated that both open and closed loop vibrotactile approaches effectively improved gait performance by addressing hypokinetic and bradykinetic aspects in PD patients. Further combination of both visual and somatosensory cues at open and closed-loop, in a novel strategy focused in addressing FoG, was validated with high acceptability by healthy participants and promising potential for motor rehabilitation in different scenarios of daily living. These findings supported its long-term application with end-users, with effectiveness depending on consistent and appropriate use of cueing. Additionally, continuous training with the strategy allowed improvements in PD patients FoG ocurrences, motor performance and quality of life, highlighting the positive impact of this strategy on daily living for PD patients. Keywords: Cueing; Freezing of Gait; longitudinal study; Parkinson’s Disease; Quality of Life vi CONTENTS 1. INTRODUCTION .................................................................................................................... 1 1.1. MOTIVATION .......................................................................................................................................... 1 1.2. PROBLEM STATEMENT .......................................................................................................................... 4 1.3. DISSERTATION GOALS AND RESEARCH QUESTIONS ............................................................................. 4 1.4. CONTRIBUTION TO KNOWLEDGE .......................................................................................................... 7 1.5. PUBLICATIONS AND ORAL COMMUNICATIONS [24] .............................................................................. 7 1.6. MANUSCRIPT OUTLINE ......................................................................................................................... 8 2. LITERATURE REVIEW ON CUEING STRATEGIES FOR PD PATIENT’S GAIT REHABILITATION... 9 2.1. METHODOLOGY..................................................................................................................................... 9 2.1.1. RESEARCH METHODOLOGY .................................................................................................. 9 2.1.2. SELECTION STRATEGY ........................................................................................................ 10 2.1.3. DATA EXTRACTION .............................................................................................................. 10 2.2. RESULTS ............................................................................................................................................. 10 2.2.1. TECHNICAL SPECIFICATIONS .............................................................................................. 11 2.2.2. CLINICAL SPECIFICATIONS .................................................................................................. 27 2.3. DISCUSSION........................................................................................................................................ 40 2.3.1. RQ1: HOW HAVE CUEING STRATEGIES BEEN USED AND APPLIED FOR TARGETING MOTOR SYMPTOMS OF PD? ............................................................................................................................ 40 2.3.2. RQ2: WHICH IS THE EFFECTIVENESS OF CUEING STRATEGIES CONCERNING MOTOR EFFECTS AND QOL? ........................................................................................................................... 42 2.4. CONCLUSIONS .................................................................................................................................... 45 3. SOLUTION OVERVIEW ......................................................................................................... 47 3.1. HARDWARE ......................................................................................................................................... 47 3.2. CONTROL AND MONITORING STRATEGIES .......................................................................................... 48 3.2.1. MOTION MONITORING ......................................................................................................... 49 3.2.2. GAIT EVENT-DRIVEN CUEING ............................................................................................... 50 3.2.3. CONTINUOUS CUEING ........................................................................................................ 51 3.2.4. FOG PREVENTION CUEING .................................................................................................. 51 3.3. DESKTOP APP (MATLAB) ..................................................................................................................... 52 3.4. CONCLUSIONS .................................................................................................................................... 55 4. CROSS-SECTIONAL STUDY.................................................................................................. 57 4.1. METHODOLOGY................................................................................................................................... 57 4.1.1. PARTICIPANTS ..................................................................................................................... 57 xiii JoLO – Judgement of Line Orientation L LEDD – Levodopa Equivalent Daily Dosage M MBEST – Mini Balance Evaluation Systems MDS-UPDRS – Movement Disorder Society-Unified Parkinson’s Disease Rating Scale MoCA - Montreal Cognitive Assessment MMSE - Mini-Mental State Examination N NFOG-Q – New Freezing of Gait Questionnaire P PD – Parkinson’s Disease PDQ-39 – Parkinson’s Disease Questionnaire PIGD – Postural Instability and Gait Disorder Q QoL – Quality of Life R RAS – Rhythmic Auditory Stimulation RCT – Randomized Controlled Trial S SD – Standard Deviation SIP – Stepping-in-place SNC – Senior Neurological Campus STAI – State-Trait Anxiety Inventory SUS – System Usability Scale T TEDD – Total Electrical Energy Delivered xiv U UMinho - University of Minho 1 1. INTRODUCTION The present report aims to outline the work established, along with respective findings, of a dissertation in the scope of the second year of the master’s program in Biomedical Engineering – Medical Eletronics field, at the University of Minho (UMinho), entitled “Motor effects of Biofeedback in Parkinson’s rehabilitation”. This dissertation was designed to be accomplished during the academic year of 2023/2024, at the Biomedical Robotic Devices Laboratory (BiRDLab) – Center for MicroElectroMechanical Systems (CMEMS), research center of UMinho. Experimental trials were also executed in the scope of this dissertation, taking place in both the Clinical Academic Center (2CA) at the Hospital of Braga, and the Senior Neurological Campus (CNS), of Braga. The work developed was supervised by a robotics professor at UMinho and principal investigator of BiRD Lab, Cristina Santos; and a Ph.D. biomedical engineer student also at BiRDLab, Cristiana Pinheiro. The project also counted with the support of a neurologist at Braga Hospital, Drª Ana Margarida Rodrigues, which partnered with the project recruiting PD patients for the clinical trials in the scope of this dissertation. 1.1. MOTIVATION Parkinson’s Disease (PD) stands as the second most prevalent neurodegenerative condition, following Alzheimer’s, impacting particularly the aging population [8], [9]. Therefore, PD becomes an exacerbating concern as the world demographics continue to shift towards an increasingly aged population. This transition not only results in an increase in the prevalence of the disease but also elevates its economic implications, ultimately posing as a substantial burden on healthcare systems worldwide [1], [9]. PD is classified as a synucleinopathy, a neurodegenerative disorder characterized by the loss of dopamine-producing neurons within the nigra region of brain [1], [2]. Although the causes of Parkinson's disease (PD) remain a challenge to ascertain, it is known that abnormal alpha-synuclein protein buildup in the substantia nigra causes dopamine-producing neurons to degenerate, thereby lowering dopamine levels [1]. The direct and indirect channels that pass through the striatum are impacted by this decrease, which in turn affects an individual's motor response and, therefore, his motor activity [1], manifesting as debilitating motor symptoms such as Freezing of Gait (FoG), bradykinesia (slowness of movement), akinesia (inability to voluntarily perform a movement), tremors (involuntary repeated rhythmic movement of a body part), rigidity (continuous involuntary muscle contraction), and impaired motor 2 reflexes [1], [8]–[10]. Even though each patient affected may experience different motor symptoms, gait disturbances limit autonomy in carrying out everyday chores. As a result, become more dependent on help and Quality of Life (QoL) declines [7]. FoG is characterized by the sudden inability of generating movement, therefore freezing in place, and is one of the main symptoms that greatly affects QoL [3]. Not all PD patients’ experience FoG, since it is a symptom that manifests in more advanced stages of the disease [11]. However, it is a frightening reality for patients to consider has the disease advances. FoG manifests spontaneously, being mainly triggered by certain tasks like turning or gait initiation, but also spatial constraint and stress situations, like staircases and passing throw narrow paths, as for example door frames [3], [10]. The episodic nature of the symptom makes it one of the most challenging ones to understand, but also among the most dangerous for patients since it can result in falls [10]. However, studies have shown that these gait impairments in PD patients are not solely motor phenomena. Instead, they may arise from an intricate interplay among sensorimotor, cognitive, and emotional factors [11], [12]. This interaction between these domains leads to the expression of a broader spectrum of symptoms in PD, including sleep disturbances (insomnia), mood disorders (anxiety and depression), cognitive impairment (attention deficits), psychosis and confusion and even autonomic dysfunction (balance impairments and postural instability) [1], [13]. A thorough comprehension and research of these symptoms in essential to discover effective therapeutic solutions focused on enhancing patients’ QoL. Despite ongoing research efforts and advances in understanding the disease, the ethology and pathophysiology of PD remains enigmatic, making its management and treatment a formidable challenge. Therefore, the need for comprehensive research to uncover the underlying mechanisms and exploring potential therapeutic pathways has never been more crucial [3]. PD remains an incurable condition, underscoring the need to develop therapeutic interventions focused on enhancing patients Quality of Life (QoL) [2], [3]. In the present days, pharmaceutical interventions with dopaminergic medication, such as drug delivery using levodopa to elevate dopamine levels in the brain, remains the primary treatment modality [4], [8]. A critical challenge in PD lies in the issues presented by certain symptoms, such as worsen FoG and inducing dyskinesias (abnormal involuntary movements) associated with dopaminergic medication usage [8], [14]. These motor fluctuations comprehend an obstacle to optimal treatment, limiting the efficacy of conventional therapies, hence, it is imperative to explore therapeutic alternatives that can mitigate these challenges. 3 Nevertheless, there is a growing interest in alternative and complementary therapies, such as Deep Brain Stimulation (DBS), occupational therapy, and physical therapy, which have gained prominence in recent years [4], [8]. However, these therapies also entail drawbacks, such as relatively high cost or long-term expenses [15]. Physical and occupational therapies, to be more effective require more experienced therapists, therefore being subjective which can also lead to slower progress, that can affect patients’ motivation [15]. Both methods by depending on therapist don’t allow patients to feel a sense of autonomy, which is also essential for motivating them to perform the treatments [16]. Some methods like DBS are invasive needing surgery, which entails associated risks. Additionally, there are patients that are not suitable for neither alternative due to the heterogeneous pathophysiology challenge of the disease, so it is necessary to give patients more options developed taking accountability of these lapses [3]. Cueing strategies are therapeutic approaches that emerge as promising alternatives which have great potential in treating PD [3], [4]. Cueing Systems (CSs) help patients executing motor tasks, providing external auditory, visual, or somatosensory cues delivered at a time or distance pace, for continuous enhancement of motor activity, dependent or independent of the end-user motor behaviour [5]. Cueing strategies can be categorized in two groups, depending on how they are delivered: open-loop or closedloop. Open-loop cueing provides cues at a marked pace, independently of the users. On the other hand, closed-loop cueing is dependent of end-user, by adapting cueing rhythm to the end-users’ behaviour [17]. Biofeedback is a type of closed-loop cueing strategies since provides the opportunity to monitor and control, offering real-time feedback, that assists patients in becoming aware and self-regulating their motor functions [7]. Both methods can be used separately and combined, allowing to monitor patients’ reactions to cueing during gait, which arises great potential since provides a customized course of treatment for the patient [7]. Cueing stimulates a different pathway then the one affected by PD. Diverse cueing strategies, such as Rhythmic Auditory Stimulation (RAS), can activate channels from different brain regions not directly correlated with dopaminergic pathway’s, including cerebellar and prefrontal channels [6]. Thus, since these alternative pathways are spared during the diseases’ progression, retain potential for effective compensation and restoring of pace and rhythm during gait[6] Wearable devices are the future of at-home or daily treatment options [7]. Motor tracking systems, such as wearable inertial sensors, allow the analysis of patients’ gait, providing insights into different gait parameters such as spatiotemporal parameters, kinematic parameters, Anticipatory Postural Adjustments (APAs) and duration and number of FoG episodes [4], [5], [7]. Closed loop cueing systems, 4 by integrating motor tracking sensors with actuators capable of delivering auditory, visual or vibrotactile cues, arise the potential for guiding and correcting gait impairments or proactively addressing motor symptoms prior to their manifestation [4], [5], [7]. Therefore, closed loop cueing systems come to the forefront as a pioneering solution within this framework [7]. 1.2. PROBLEM STATEMENT Despite the availability already of existing therapeutics and technological solutions for treating Parkinson, it is imperative to explore novel options that can enhance patients QoL [3]. Therefore, longterm wearable devices are rising as a potential solution for a more convenient, low-cost, and portable athome therapy [7]. However, it is essential to study and evaluate their efficacy and their impact on the disease to determine their viability as a potential treatment, since the training for daily tasks remains a challenge [7]. There are still gaps that need to be filled, regarding some cueing strategies, like vibrotactile cues [18]. Despite different types of cueing strategies already being tested, there is lack of information about: the types of cueing, and control strategies, that have better results in treating gait impairments of PD patients; and the long-term effects in enhancing PD patients motor effects and thus QoL [7], [18]. Since some cueing strategies, like vibrotactile cues, yield limited results also since benchmark has not been accomplished, resulting in lack of conclusions about their efficiency in treating PD symptoms, it is important to explore the potential of this type of cueing [18], [19]. Therefore, in this project it will be analysed the potential of the +sensBand, a wearable device, developed in BiRDLab, that uses vibrotactile cueing to enhance PD patients’ rehabilitation [19]. To fully comprehend the potential of cueing strategies, it is crucial that clinical trials are conducted [7]. However, relying solely on single-use trials proves insufficient, given the abundance of existing results. The next step is to conduct long-term studies, designed to assess the viability of integrating these strategies for ongoing utilization and into daily tasks [7]. Only through such means will it be possible to innovate and start impacting patients QoL. 1.3. DISSERTATION GOALS AND RESEARCH QUESTIONS The main goal of this dissertation is the development and consequent analysis of a longitudinal clinical study of Cueing strategies in PD patients, evaluating the long-term use motor and QoL impact and potential of wearable cueing systems, using +sensBand. The +sensBand is a plug-and-play wearable 5 cueing system created by BiRDLab. It integrates an inertial sensor and a miniaturized camera, to track and monitor users’ movements, and features vibration sensors that provide somatosensory cueing, aiming to counteract any symptoms that may arise during gait tasks, such as FoG. To achieve the main goal, several sub-objectives have been defined, with some respective key performance indicators (KPI’s), to accomplish gradually along this dissertation: • Objective 1: Literature review and analysis focused on Parkinson clinical study, using cueing strategies. This review will investigate how cueing strategies have been used (auditory, somatosensory, and visual) and applied (single use sessions or longitudinal studies) for combating motor symptoms of PD and their effectiveness (e.g., effects over clinical scales, APAs, step and stride length, cadence, velocity, number of FoG episodes). A clear summary of the main cueing strategies implemented, specifications and respective outcomes reviewed should be accomplished (KPI 1). • Objective 2: Familiarization with the +sensBand system and examination of control strategies suitable for long-term clinical trials with Parkinson. Throughout the project, the intention is to leverage the pre-existing technology, +sensBand, and subsequently analyse its potential, alongside novel cueing strategies developed based on the conducted literature review. Therefore, it is imperative to consider an acquainting phase dedicated to understanding the device and its operation mode. Moreover, familiarization will include the participation on the conduction and data analysis of a currently in course cross-section clinical study with +sensBand in Parkinson. As for KPI 2, it included the summary of the different control strategies of +sensBand that should be considered for the further protocol development and included also data collection of 15 more patients gait metrics to complete the on-going cross-sectional study and its analysis. • Objective 3: Development and validation of a novel cueing strategy adequate to target FoG. This strategy will be designed for PD patients in more advanced stages of the disease, based on the strategies already developed and literature conclusions, allowing a more complete and intense strategy targeting FoG. It is expected to perform a cross-sectional study, addressing the usability and workload of the strategy, to grant that should be suitable for use in scenarios of daily living. This objective will be measured by KPI 3: (i) implementation of a novel cueing strategy to assist PD patients with FoG; 6 (ii) implementation of both visual and somatosensory cues simultaneously; and (iii) granting that the strategy does not overstimulate patients with SUS scores above 65 points and according to the respective individual items reference benchmarks [20]. • Objective 4: Development, validation and conduction of the proposed protocol with end-users. This protocol will be designed for PD patients, based on the literature conclusions, functionalities of +sensBand, and by setting the necessary tradeoffs with clinicians. It is expected to perform a longitudinal study, addressing the limitations but also potential of motor and QoL effects of long-term cueing use. This objective will be measured by the development of a continuous study circuit that will integrate different motor tasks susceptible of inducing FoG. To evaluate the different aspects of the device, it is imperative to perform the multiple training sessions (longitudinal study) using the +sensBand device and include evaluation time points for acquiring sensor based and clinical data, assessing the motor (FoG episodes and spatiotemporal parameters) and QoL effects of the training. During the protocol conduction it should be accomplished a reduction of patient’s FoG episodes (KPI 4). • Objective 5: Data analysis and discussion regarding the motor and QoL effects of long-term cueing use in Parkinson. The data collected from the protocol conduction should be examined and discussed, to draw conclusions about the cueing long-term use effects over PD patients’ symptoms. This goal will be measured considering KPI 5, indicating that it should be accomplished spatiotemporal differences of at least 30% between study groups [21], PDQ-39 score minimal clinical improvements of at least 4.72 points [22], and minimal NFOG-Q score changes of at least 3 points [23]. • Objective 6: Disseminate research by writing and presenting dissertation document, and articles (KPI6). Furthermore, to encourage the research for the literature review, this dissertation aimed to answer the following queries: • Research Question 1: How have cueing strategies been used and applied for targeting motor symptoms of PD? 7 • Research Question 2: Which is the effectiveness of cueing strategies concerning motor effects and QoL? • Research Question 3: How does +sensBand cueing system long-term use affect Parkinson’s motor symptoms and QoL? 1.4. CONTRIBUTION TO KNOWLEDGE The work done during this dissertation contributes with novel cueing solutions based on wearable devices to improve motor symptoms of PD patients. The main contributions outlined are: ❖ A literature review about clinical studies using sensory cueing solutions for PD motor rehabilitation, compiling information about: different cueing devices used, cueing modalities used and the way they were delivered, sensors used for data collection, sample sizes and study groups considered, protocol duration, training dosage and outcomes measured (Chapter 2, Objective 1). ❖ Finding how cueing strategies have been applied to target PD motor symptoms and their effectiveness regarding those effects but also QoL (Chapter 2, Objective 1). ❖ Implementation of a novel sensory cueing strategy (FoG prevention cueing) integrated on the +sensBand device designed to target FoG in PD patients (Chapter 3, Objective 3). ❖ Improvement of an ongoing study, resulting in a journal paper to be published, about the efficiency of using different somatosensory strategies in PD patients, comparing the efficiency of continuous rhythmic vibration with vibrations delivered at a personalized rate according to patients' final contacts of the right foot (Chapter 4, Objective 2). ❖ Validation of the usability and workload of the novel sensory cueing strategy, resulting in the publication of a conference paper (Chapter 5, Objective 3) and further validation of the strategy testing in PD patients (Chapter 6, Objective 3). ❖ A longitudinal study about the use of the novel cueing strategy in PD patients for motor rehabilitation and QoL improvements, regarding the long-term effects as a potential solution for mitigating FoG (Chapter 6, Objectives 4 and 5). 1.5. PUBLICATIONS AND ORAL COMMUNICATIONS [24] This dissertation resulted in two articles: one conference paper as first author entitled “Robotic sensory cueing during gait: a usability study with scenarios of daily living”, IEEE International Conference 8 on Autonomous Robot Systems and Competitions, Paredes de Coura, Portugal, 2024, Maria Pimenta, Cristiana Pinheiro, Cristina P. Santos [24]; and one journal paper, as first author, currently under revision entitled “Gait event-driven biofeedback to improve gait impairments in Parkinson’s disease”, Movement Disorders, Official Journal of the International Parkinson and Movement Disorder Society (impact factor of 7.4), 2024, Maria Pimenta, Cristiana Pinheiro, Helena R. Gonçalves, Ana Margarida Rodrigues, Cristina P. Santos. Following the “Robotic sensory cueing during gait: a usability study with scenarios of daily living”[24], besides the published conference article, this work also resulted in an oral presentation during the ICARSC 2024 conference. 1.6. MANUSCRIPT OUTLINE This document is organized into seven chapters. Chapter 1 presents the motivation for this dissertation, as well as the problem statement, objectives to accomplish, the contribution of this work for further research and the resulting publications and presentations. In the following chapter, Chapter 2, the state of art is outlined presenting the literature review found until the present moment. Chapter 3 presents a conceptual overview of the main components, hardware and software, used during this dissertation. Chapters 4, 5 and 6 contain the description and results of the three different studies done, respectively: a cross-sectional study that compared different cueing strategies in PD patients; a usability and mental workload study of a novel cueing strategy performed with healthy participants; and a longitudinal study using this novel strategy in PD patients with FoG. Lastly, Chapter 7 resembles the conclusions of all the work done for this project. 15 Studies (year) Sensory cues Gait assessment sensors Device type Device description Location Mode Device Location Parameter Ginis et al. (2016)[51 ] CuPiD: audiobiofeedback (ABF-gait app) and FOG training (FOG-cue app) using a smartphone (Galaxy S3-mini, Samsung, South Korea), Earphones or smartphone speaker 2 IMUs (EXLs3, Italy) Shoes using ABF app, and above ankles using FOG app Spatiotemporal (velocity, cadence, stride-length) Clinical (2-min Walk Test, MBEST, FSST, FES-I, PASE) Tosseram s et al. (2022)[52 ] Metronome NM 3-dimensional motion capture analysis (VICON from Oxford Metrics, UK) 16 markers placed following the Plug-in Gait Lower Body Model Spatiotemporal (gait variability, stride time) Capato et al. (2020)[53 ] Speaker (JBL Go Portable Wireless Speaker): metronome sound NM Video cameras NM Clinical (MBEST, TUG dual task performance, UPDRS) Chang et al. (2019)[54 ] Metronome NM EMG through surface eletrodes Tibialis anterior muscle Physiological (MEP, CSP, AMEP, SICI, ICF) Kistler force plate (Kistler AG, Switzerland) On the floor Spatiotemporal (SIP performance) GAITRite Analysis System Pathway Spatiotemporal (gait speed, variability, cadence, and stride length) LiraniSilva et al. (2017)[55 ] Noneletronic Textured insoles: with semiesphered elevations Distal phalanx of the hallux, heads of metatarsophalangeal joints and heel optoelectronic tridimensional system (Nothern Digital Inc.) 4 markers: 5th right and 1st left metatarsal joints and lateral face of the right calcaneus and medial face of the left calcaneus Spatiotemporal (stride length, duration and velocity, cadence, step width, stance phase duration and % of double-support phase duration) Schlenste dt et al. (2020)[56 ] Eletronic VibroGait Wrists 3 Opal IMUs (APDM Inc., USA) 1 at posterior trunk (L5) and 1 at each shin Kinematic (APAs) Surface eletromyography (EMG, Wave by Cometa) Tibilias anterior and gastrocnemius mediali bilaterally Physiological (TFL muscle activity) Rosenthal et al. (2018)[57 ] CueStim (custom-built electrical stimulator): worn on the waist with skin surface eletrods Hamstrings or quadriceps of body side most affected NM (recordings) NM Spatiotemporal (time of completing walking task) Clinical (VAS, FPR scale, number of FoG episodes) Peppe et al. (2019)[58 ] Equistasi®: wearable focal mechanical vibratory device C7 vertebra and gastrocnemius 6-cameras stereophotogrammetric system synchronized to 8-channel sEMG system 4 muscles bilaterally: Rectus Femoris, Tibialis Anterior, Biceps Femoris and Gastrocnemius Lateralis Physiological (Peak of the envelope and occurrence) ElTamawy et al. Vibratory devices (VDs) (OPTEC Co. Ltd., Japan) Into-shoes Qualysis ProReflex motion analysis system (Qualysis Medical AB, Sweden) NM Spatiotemporal (cadence, stride length) Kinematic (hip, knee, and ankle joints’ angular excursion) 16 Studies (year) Sensory cues Gait assessment sensors Device type Device description Location Mode Device Location Parameter (2012)[59 ] Zhao et al. (2016)[60 ] Google Glass (Explorer version 2, XE 22.0, Android 4.0) Headset + 7 IMU’s (MVN motion capture suit (Xsens, Netherlands)) Pelvis, upper legs, lower legs and feet Spatiotemporal (velocity, cadence, stride length and variability) Vídeo cameras Start and midpoint of walkway Clinical (presence and severity of FoG) Carpinell a et al. (2017)[61 ] Gamepad: 6 wearable inertial sensors (TMA) and PC with LCD screen (16'') NM 6 IMUs (TMA) Upper and lower trunk and lower limbs through elastic belts Spatiotemporal (velocity, balance (BBS), 10-min walk test) YogevSeligmann et al. (2023)[62 ] Self Pro system: computer, projector and spatial camera (MS Kinect Azure camera & Software Developer Kit; SDK) Projected light stripes and metronome beats Placed at 2m from the walking area (stripes projected at 110% of the user’s step length) Spatial camera NM Spatiotemporal (step length and time) Lee et al. (2023)[63 ] Google glass: custom-made auditory-visual cue applications "Walk With Me" and "Unfreeze Me" Headset NM (videotaped) NM Spatiotemporal (time to complete task) Janssen et al. (2017)[64 ] Smartglasses Headset Vídeo cameras NM Clinical (number and duration of FoG episodes) Metronome NM 17 IMU’s (XSens, Netherlands)) Feet, lower legs, upper legs, pélvis, hands, forearms, upper arms, sternum, shoulders and head Spatiotemporal (stride length, time, cadence and velocity) Yang et al. (2022)[65 ] Wireless-controlled commercial earphones: music with different rhythms Ears Smart insoles: with 6 PSUs 1 PSU set at the heel and 5 at the forefoot (3 at metatarsal heads and 2 to detect foot-inversion and foot-eversion) Spatiotemporal (velocity, variability and stride length) Custom-made laser-light generator Waist or top of shoes, projects line in front of patients Jiang et al. (2006)[66 ] Eletronic Piezo buzzer connected to a stimulator NM Optotrak 3020 system (Northern Digital Inc., Canada): with 3 IREDs: 1 to each heel and 1 over the zygomatic bone. Spatiotemporal (step time and length, velocity) Noneletronic Stripe tape lines Placed at patients step length or 40% of their height Forceplate (Advanced Mechanical Technology Inc., USA) Where patients stood Kinetic (timing and magnitude of weight shift and push-off force) Lee et al. (2012)[67 ] Eletronic Metronome NM 3-dimensional computerized motion analyzer (Motion Analysis; CA) 19 markers: lower limbs and trunk, enabling 3D analysis of the foot, shank, thigh, pelvis, and trunk body segments Spatiotemporal (velocity, cadence, stride length, step length, and doubleand single-limb standing time); Kinematic (degrees of maximum pelvic tilt, hip flexion, knee flexion, and ankle dorsiflexion in the sagittal plane) Clinical (frequency of FoG episodes, total step numbers, and total time per walking cycle) Noneletronic Stripe tape lines Placed at normalized step lengths distance 17 Studies (year) Sensory cues Gait assessment sensors Device type Device description Location Mode Device Location Parameter Sijobert et al. (2017)[68 ] Eletronic Eletrostimulator Feet: anode (dorsum) cathode (arch) Video cameras NM Clinical (number and duration of FoG episodes) Laser cane Projected horizontally from cane tip on the floor Mancini et al. (2018)[69 ] Speakers NM 8 Opal IMUs (APDM Inc., USA) Both shins, feet, wrists, on the sternum and posterior trunk (around L5) Clinical (duration of FoG and FoG-ratio, smoothness) VibroGait vibration sensors Wrist Klaver et al. (2023)[70 ] Metronome (Natural metronome App) NM Vídeo camera NM Clinical (number of FoG episodes and duration) Vibrating socks: flat mini vibration motor (Adafruit Mini Motor Disc 1201) Below arch of both feet 17 IMU’s (Xsens, Netherlands) NM Spatiotemporal (velocity and variability) SilvaBatista et al. (2022)[71 ] Speaker: delivered metronome beat Center of laboratory 8 IMU’s (APDM Inc., USA) Both shins, feet, wrists, on the sternum and posterior trunk (around L5) Clinical (FoG index) Vibrogait Both wrists Lu et al. (2017)[72 ] Vertical array of light emiting diodes (CREE) 3m in front of participants 2 force plates (9260AA; Kistler) On the floor, where patient starts task Kinetic (ground reaction forces and CoP) Kinematic (APA’s timing and amplitude) Vibrotactile transducer (C-3 Tactor, Engineering) Ankle (lateral malleolus) of initial stance leg Speakers (HPG-100N, Atlas Sound) NM McCandle ss et al. (2016)[73 ] Metronome (Peterson BodyBeat Pulsing Metronome) Belt, placed over right side of the pelvis + 10 camera Qualisys motion analysis system 13 markers: feet, shanks, thighs, pelvis, trunk, head, arms and forearms Clinical (number of FoG episodes) Spatiotemporal (first and 2nd step length) Laser cane (Laser Cane, U-Step Projected line on the floor 4 force plates (Advanced Mechanical Technology, Inc.) NM Kinetic (forward and sideways CoM velocity; forward and backward sways and number of sideways sways) Nieuwboer et al. Light emitting diode: flash On glasses Vitaport Activity Monitor (TEMEC Instruments Inc, Netherlands), with 5 accelerometers 1 on each leg (lateral thigh), 3 placed on the lower third of the sternum orientated Spatiotemporal (turning speed and time during task) 18 IMUs: Inertial measurement units; EEG: Eletroencephalografy; NM: not mentioned; TUG: Timed Up an Go; EMG: eletromiography; ML APAs: Medio lateral Antero posterial adjustments; LED: Light emmiting diode; VDs: Vibratory devices; PSU: Pressure sensitive units; CoM: Center of Mass; CoP: Center of Pressure; IREDs: Infrareds; : Visual cues; : Auditory cues; : Somatosensory cues. Studies (year) Sensory cues Gait assessment sensors Device type Device description Location Mode Device Location Parameter (2009)[74 ] Audio tone Via earphone in the sagittal, longitudinal, and transverse planes. Miniature cillinder: pulsed vibration Wristband Suputtitad a et al. (2022)[75 ] Waistbelt cueing device LED laser Projected line in front of patient 2m RS foot scan Embedded in the center of walkway Spatiotemporal (velocity, stride length, and cadence) Eletronic buzzer NM Vibration motor Waist 19 SENSORY CUES This literature review intends to explore cueing strategies in PD patients. Therefore, one main topic to consider is the different devices used for cueing. By examining the present studies, it was possible to split the cueing devices in 2 categories: eletronic, considered when mentioned the use of at least an eletronic device providing the cueing mechanism (43 studies) [32]–[54], [56]–[75]; and non-eletronic, when it is not mentioned the device in itself used or the cue does not need a robotic device to be provided (14 studies) [26]–[32], [43]–[45], [52], [55], [66], [67]. It is worth noting that 7 studies, [32], [43]–[45], [52], [66], [67], used both types of cueing devices, one eletronic and another non-eletronic, and compared the motor effects of each cueing device in PD patients. Cueing strategies can also be organized by the different cueing mode used. Essentially, there are 3 types of cueing modes: visual cues, auditory cues, and somatosensory cues; depending on the sense they act on, and it is possible to actuate using one single cueing mode or different modes combined. Visual cueing was the most used mode (33 studies) [26]–[45], [60]–[68], [72]–[75]; followed by auditory cueing (24 studies) [46]–[54], [60]–[67], [69]–[75]; and lastly somatosensory cueing (13 studies) [55]– [59], [68]–[75]. However, four different combinations were found within this studies, being the most common testing both Auditory and Visual cueing (8 studies) [60]–[67]; followed by testing all three Auditory, Visual and Somatosensory cueing (4 studies) [72]–[75], then testing Auditory and Somatosensory Cueing (3 studies) [69]–[71] and lastly Somatosensory and Visual cueing (1 study) [68]. Of the non-eletronic devices used for cueing, all studies used visual cues [26]–[32], [43]–[45], [66], [67] except one study that used somatosensory cueing [55] and another that used auditory cueing [52]. The non-eletronic visual cueing devices consisted of tape lines (8 studies) [27], [28], [30], [31], [43], [44], [66], [67] or tape stars arrangements (1 study) [26] to visually help patients perform the gait task by spatially marking the steps or the turning positions; checkered floor patterns to assist patients floor visualization, intending for the squares to assist foot placement during gait (1 study) [32], objects such as wooden rods to give a 3D spatial sensation of step positioning (1 study) [29], and an inverted wooden walking stick, with a slat of wood attached at the bottom to give a step line cue for spatial guidance and support, for patients when having FoG episodes (1 study) [45]. As for the non-eletronic somatosensory cueing device, Lirani-silva et al. [55], used textured insoles with 9mm of diameter sphere elevations, located under the distal phalanx of the hallux, heads of metatarsophalangeal joints and heel, to study the proprioceptive plantar effects in gait. Lastly, Tosserams et al. [52] used an internal cueing 20 technique that consisted of the patient self-counting and matching each step to the counting rhythm as a non-eletronic cueing method for gait rehabilitation. As for the eletronic devices, 22 studies used auditory cues [46]–[51], [53], [54], [60]–[65], [67], [69]–[75], 21 studies used visual cues [33]–[42], [60]–[65], [68], [72]–[75] and 12 studies used somatosensory cues [56]–[59], [68]–[75]. Of that, 6 studies used both visual and auditory cueing electronic devices [60]–[65], 4 studies used all 3 modes of cueing [72]–[75], 3 studies used both somatosensory and auditory cue modes [69]–[71], and 1 used both visual and somatosensory cueing [68]. AUDITORY CUEING The reviewed auditory cueing electronic devices consisted mainly of metronomes (11 studies) [47], [49], [52], [54], [62], [67], [70], [73], [53], [69], [71]to give temporal rhythmic input in PD patients by making them match each step to the metronome beats. Similarly, and with the same purpose, 3 studies used smart glasses with auditory rhythmic cues, 1 using Cinoptics (Cinoptics, Maastricht, the Netherlands) [64], and 2 using Google Glass (Explorer version 2, XE 22.0, Android 4.0) [60], [63]; 2 studies used buzzers that produced rhythmic beeps, 1 using a Piezo buzzer connected to a stimulator [66] and another using an electronic buzzer [75]; and 1 study used a rhythmic auditory tone delivered via earphones [74]. All auditory cueing devices reviewed delivered open-loop cueing. As for other auditory cueing strategies used, 1 study used a piano arrangement of children’s song “Row, Row, Row, Your boat”, providing rhythm to help set the pace of the stride [46]; 1 study used speakers to delivered diverse timed tones [72]; other study used an iPod Touch playing music or podcasts as a reward when stepping correctly [50]; other used a Gamepad system that produced a sound alarm to guide patients when doing a wrong movement [61]; and lastly one study used a smartphone based system, called CuPiD, with an audiofeedback App (ABF-gait App) and an instrumented cueing App for FoG training (FOG-cue App) to give different auditory stimuli [51]. Although Horin et al. [46] being rather focused on the song choice and characteristics of the song to be used as a cueing strategy, it was considered that it used an electronic device, such as speakers or a computer, to provide the music arrangement. Lu et al. [72] used speakers (HPG-100N, Atlas Sound) to deliver a warning and a go tone. The warning tone was set at 80dB and 1000Hz, and the go tone set at 90dB and 2000Hz, and patients were instructed to start walking only after hearing the go tone. The tones were delivered using 3 different timing protocols: a fixed delay; a countdown and a random delay. The countdown consisted of giving 3 warning 21 tones followed by a go tone, each 1s apart, giving patients the sensation of counting 3-2-1-go. The fixed delay was based on a similar principle but instead gave a warning cue followed 3s later with a go tone. As for the random delay cue, a warning cue was played and 4 to 12s later a go cue followed. Different timings were studied to analyse if had any implications in gait initiation for PD patients. Note that in Chomiak et al. [50] an iPod Touch was strapped in patients’ knee and paired to an earphones’ headset. Patients were divided in two groups, one that received music stimuli and other that received podcast stimuli. In order to maintain the music or podcast playing, patients had to perform a stepping-in-place (SIP) task correctly, which means essentially, they had to stand on one leg while lifting the other. In this study the music and podcast cues were given as a reward for correct task performance, however since music provides rhythmic input, and podcasts provide information that need more of patients’ attention, the two groups with different auditory stimuli were analysed and compared to understand the different impact in gait training. VISUAL CUEING The reviewed visual cueing eletronic devices included: laser beam projection devices (15 studies), such as waist belts [43], [75], shoes [33], [40], laser canes [36], [45], [68], [73], wheeled laser walkers [35], [37], [38], laser ankle bracelets [39] or just laser generators [34], [44], [65] attached at different positions such as waist, chest or feet to project guide lines on the walking path, to help patients visualize each step position; smartglasses (4 studies) such as Cinoptics [64], Google Glass [60], [63] or HoloLens [41], providing augmented and mixed reality cues like stripe lines, warning light flashes, footsteps, 3D bars and staircases to analyse the effects of FoG inducing ambients and spatial awareness cues in mitigating FoG, and also comparing the effects of auditory and visual stimuli combined for better spatial and temporal orientation [60], [63], [64]; projectors (2 studies) such as Self Pro system [62], used to project stripe lines for stepping, or 3 Digital Light Processing projectors [32], used to project different checkered patterns on the floor of the walking path, and compare the motor effects in PD patients of this type of visual cue with the same non-eletronic cueing strategy; and light emitting diodes (2 studies) providing a vertical array of lights [72] or attached to patients’ glasses [74], to give warning cues for gait initiation or rhythmic flashes of light for step rhythmicity during walking. Other methods such as a Gamepad [61] or an iPad [42] were also used. It is important to note that only 8 studies used closed-loop visual cueing strategies [33], [34], [37]–[39], [42], [45], [61], 3 of which compare open and closed-loop visual cueing modalities [34], [38], [45]. 22 Barthel et al. [33], uses laser shoes as the cueing mechanism, to provide laser lines in front of patients during gait. The laser beam is activated by a switch located under the sole of the contralateral foot, therefore activating the laser during heel strike of the opposite foot, guiding the lifted foot where to step. When the foot is lifted laser is deactivated while simultaneously the opposite foot, now on the ground, activates the opposite laser beam. This cycle repeats itself during the gait cycle, providing cues only during certain gait-events (heel-strike), therefore using closed-loop mode. Ratanasutiranont et al. [39], similarly provided laser lines automatically at closed loop, but instead of using laser shoes it used an ankle bracelet with an inertial sensor to detect patients walking pattern and provide the laser line following each step. Carpinella et al. [61] used a Gamepad, that combined both auditory and visual cueing, to give biofeedback during different motor tasks for patients to correct their movements. The Gamepad integrated an LCD screen that allowed patients to see bars filling or cartoons mirroring their movements, and when doing an incorrect movement (which means a movement bellow or above a certain threshold) the screen gave a red warning, a warning tone, or both, to indicate patients to repeat the movement correctly. Fig. 2 shows examples of the exercises included in the Gamepad. Fig. 2. Examples of Gamepad visual cueing exercises [61]. Similarly, Byl et al. [42] uses an iPad to display foot pressing indicators in green, yellow or red, during different mobility tasks, to let patients aware of their gait performance and correct their movements. Lastly, as for the 3 studies that use both open and closed loop cueing, Tang et al. [34] compares laser lines provided continuously, i.e. fixed projection on the ground during walking (open-loop), with provided rhythmically, i.e. personalizing the frequency of laser lines projection according to patients stride cadence (closed-loop). Cubo et al. [38] and Kompoliti et al. [45] both used similar methods by comparing wheeled walker and laser cane, respectively, projection of laser lines continuously during gait (open-loop) with provided only when patients choose to, via switches on the handles that allowed them to activate and deactivate the cues (closed-loop). Bunting-Perry et al. [37], although not comparing with open-loop cue delivery, it used the same handle switch mechanism to provide closed-loop laser lines during gait. 23 SOMATOSENSORY CUEING As for the somatosensory cueing devices, there can be essentially two types of devices: vibratory sensors or electrostimulation sensors. Therefore, 10 of the reviewed studies implemented vibratory systems, more specifically opted to provide closed-loop cueing using systems such as VibroGait [56], [69], [71], and vibratory socks [70]; or opted for open-loop cue delivery using devices such as a metronome in a vibratory mode [73], Equistasi [58], or other open-loop vibratory motors and transducers [59], [72], [74], [75]. Two studies opted for electrostimulation, using a wirelessly programmable electrostimulator (Phenix Neo Usb) [68] to deliver closed-loop cueing, or a custom-built electrical stimulator named cueStim [57] providing open-loop stimulation. As for the locations to position the devices to provide somatosensory cues were mainly applied at the wrists (4 studies) [56], [69], [71], [74]; waist (2 studies) [73], [75]; legs (2 studies) [57], [58]; feet (2 studies) [68], [70]; ankles (2 studies) [59], [72]; and neck (1 study) [58]. VibroGait is a vibrotactile cueing system, that integrates an inertial movement sensor (Opal IMUs), a controller unit, and a tactor unit, as shown in Fig. 3 a) [76]. Those components allow VibroGait to generate vibration on the user when detecting stance phase, which is a phase that usually can trigger FoG episodes. The actuator parts of the device consist only in the tactor unit since it is the part responsible for establishing the vibratory cue. This unit is placed on the wrists, where the vibration is applied [56], [69], following the setup shown in Fig. 3 b). The tactor unity, for optimal sensory cueing vibrates at 200300 Hz range, similarly to the vibration of a cell phone in vibration mode [69]. Fig. 3. VibroGait System characteristics [26]. Peppe et al. [58] used a different vibratory approach, with a wearable proprioceptive device based on focal mechanical vibration named Equistasi. Equistasi is made of nanotechnological fibers sensitive to temperature, able to transform body thermal energy into mechanical energy as a form of high frequency vibration. The Equistasi patch was placed in the C7, as shown in Fig. 4, and in the gastrocnemius, bilaterally, and as soon as it contacted with patients skin it started vibrating. Therefore, Equistasi provided open-loop cueing. a) Components b) Experimental setup 24 Fig. 4. Vertebral placement of Equistasi taken from [77]. Only two studies used electrical stimulation as somatosensory cueing strategy. Sijobert et al. [68] used a wirelessly programmable eletro-stimulator (Phenix Neo Usb, Montpellier, France). This stimulator was placed around the shank, and two eletrodes, an anode and a cathode, were placed on the dorsum and arch of the foot, respectively, where the stimulation was applied. The stimulation consisted of a closed-loop cueing pattern of five 500 ms/phase pulses at 200 Hz, repeated four times at 10 Hz. An IMU (HikoB) placed under the lateral malleolus on the sagital plane, was used to detect spatiotemporal outcomes and monitor gait events to provide the stimulation. When detecting heel-off, the stimulation was triggered and at swing phase stopped [68]. As for cueStim, it is a wearable two-channel electrical stimulator device, wireless controlled, that allows to monitor gait and FoG episodes in PD patients by using closed-loop electro stimulatory cues [57], [78]. The device is used as a waist belt, and integrates skin surface electrodes, placed in the hamstrings or quadriceps muscles of the body side most affected by the disease, as shown in Fig. 5 a) [57], [78]. The electrodes are responsible for inducing the electrostimulation. By controlling the voltage, a continuous series of electric bursts act on the muscles, delivering somatosensory cues for motor rehabilitation adjusted for each patient gait [57]. Each burst was delivered with 100ms of ramp-up time, then 500ms ON stimulation, followed by another 100ms of ramp-down time [57]. There was no interval following each burst, therefore OFF time was of 0ms [57]. Fig. 5 b) represents the cueing delivery strategies during gait [57], [78]. Fig. 5. cueStim System charatheristics [22], [27]. a) Experimental setup b) Cueing delivery strategy during gait 31 Studies (year) Sample Size (PD, HC) Training Dosing Motor task Evaluation Time points Medication State Outcomes On-target cueing effects Off-target cueing effects Mancini et al. (2018)[69] 43(43*,0) *25 PD+FoG; 18 PD-FoG 3 hours assessment (repeated 3 blocks of the same task under 3 different conditions) Turning While performing the different tasks, with a 10-15min break through each task Off Better turning smoothness; Number of FoG episodes. turning velocity and number of turns. Spildooren et al. (2017)[47] 29 (15*,14) *15 PD+FoG 6 trials per condition (3 turning left and 3 right) During motor task Off Number of FoG episodes; Improved head rotation and head-pelvis separation. Axial movement deficits did not improve. Das et al. (2022)[26] 63 (43*, 20) *22 PD+FoG; 21 PD-FoG 2h session During motor task On turning velocity (except PD+FOG). - Tang et al. (2017)[34] 23 (23*,0) *23 PD+FoG Few trials with 2-day interval between each condition During motor task On Rhytmic laser cue: step time, turning time and gait arrhythmicity; Number of FoG episodes. - Zhao et al. (2016)[60] 12 (12,0) Each condition tested twice for each turning task During motor task Off More stable gait pattern; Metronome was more effective and most preferred; Positive usability feedback of Google Glass. - Willems et al. (2007)[49] 28 (19*,9) *9 PD+FoG; 10 PD-FoG 3 trials per condition During motor task On - Non-significant effects (similar results to noncued) Nieuwboer et al. (2009)[74] 133 (133*,0) *68 PD+FoG; 65 PD-FoG 8 trials per condition During motor task On Auditory cueing was more effective; Improved turn times. - Silva-Batista et al. (2022)[71] 13 (13,0) 2 rsfMRI scans consisting of 10 minutes During motor task Off FoG Index improvements; Improved turning with both open and closed-loop cueing. - Holmes et al. (2015)[35] 6 (6*,0) *6 PD+FoG 45-min session with 3 trials per condition TUG During motor task On Improved walker positioning. - Stuart et al. (2021)[27] 63 (43*,20) *22 PD+FoG; 21 PD-FoG 2h session Gait initiation and Walking During motor task On gait velocity and stride length; power spectral densities over sensory regions. - Sijobert et al. (2017)[68] 13 (13*,0) *9 PD+FoG; 4 PD-FoG 5 trials per condition (3 last trials were analysed) During motor tasks NM time to complete task; Number of FoG episodes. - Bryant et al. (2010)[36] 7 (7*,0) *7 PD+FoG Rest periods between tasks Walking and Turning During motor task On and Off Green light: Number of FoG episodes; - 32 Studies (year) Sample Size (PD, HC) Training Dosing Motor task Evaluation Time points Medication State Outcomes On-target cueing effects Off-target cueing effects Turn time and number of steps. Ratanasutiranon t et al. (2023)[39] 4 (4*,0) *4 PD+FoG 2 sessions of TUG and walking; 10 min washout period between walking with laser on and off Walking and TUG Baseline; Tested again 1 week after On stride length and velocity; TUG test time; % FoG time. - Klaver et al. (2023)[70] 31 (31*,0) *31 PD+FoG 2 sessions on 2 separate days (1 On and other Off medication) Walking (passing doorway) and Turning During motor task On and Off >10% improvement in FOG; User satisfaction. Closed-loop and open-loop tactile, and auditory cueing did not significantly impact the % time and number of FoG episodes, or spatiotemporal parameters Lee et al. (2023)[63] 10 (10*,0) *10 PD+FoG 3 trials per task During motor task On “Walk with me” program: Better straight walk, dual task, and walk through doorway. “Unfreeze me” program: Only dual task improved. UPDRS III did not had significant improvements; “Walk with me” program: Worse 180 degree turning. An et al. (2023)[40] 10 (10*,0) *10 PD+FoG 10 min familiarization; 3 trials per condition; 10 min break in between tasks Gait initiation; Walking and Turning During motor task On Number, duration and proportion of FoG; Greater uptake of FP-CIT. - Longitudinal Lirani-Silva et al. (2017)[55][50] 19 (19**,0) **10 textured insole; 9 conventional insole 5 trials without insoles Walking Pre-test; After 1 week wearing the groupspecific insoles (Post-test); After 1 week wearing conventional insoles (follow-up). On Post-test textured insoles: stride length and plantar sensation; Follow-up: Only plantar sensation improvements maintained. - El-Tamawy et al. (2012)[59] 30 (30,0) 45-min session of low intensity physiotherapy program, 3 times a week, for 8 weeks; 5 min treadmill training at baseline and 25 min at the end of treatment Before and after treatment On Improved lower limb joints’ angles; Cadence and stride length. - Ginis et al. (2016)[51] 40 (40**,0) **22 CuPiD; 18 CG walk at least 3 times per week for 30 min, during 6 weeks FoG: additional 30 min, 3 times a week using the FOG-cue app baseline; post-test (after 6 weeks); follow-up (following 4 weeks) On Post-test: MBEST score; Follow-up: Maintained SF-36 (while controls decreased). MBEST not maintained at follow-up. 33 PD: Parkinson’s Disease; HC: healthy controls; PD+FoG: PD patients with FoG; PD-FoG: PD patients without FoG; EG: experimental group; CG: Control group; ML APAs: Medio Lateral Antero Posterial Adjustments; MDS-UPDRS: Movement Disorder Society – Unified Parkinsons Disease; MoCA: Montreal Cognitive Assessment; % time FoG: percent time spent frozen; TUG: Timed Up an Go; EMG: eletromiography; CoM: Center of Mass; MBEST: Mini Balance Evaluation Systems test; CoP: Center of Pressure; FOG-Q: Freezing of Gait questionnaire Studies (year) Sample Size (PD, HC) Training Dosing Motor task Evaluation Time points Medication State Outcomes On-target cueing effects Off-target cueing effects Peppe et al. (2019)[58] 20 (20,0) 8 week training Motor Balance before and after first 8 weeks; before and after following 8 weeks On Stance and double support time; Velocity and stride length; Severe patients ( H&Y) had better improvements in MDS-UPDRS III; Placebo showed some improvements in MDSUPDRS III. - Carpinella et al. (2017)[61] 42 (42**,0) **22 EG; 20 CG 20 sessions of 45 min each, 3 per week baseline; post-training; 1 month follow-up On Post-training: Better balance performance; CoP ML. At 1-month improvements were not maintained. Capato et al. (2020)[53] 154 (154*,0) 56 RAS; 50 without RAS; 48 educational program 10 sessions of 45 min; 2 sessions per week, over 5 weeks baseline (14 days prior to training); one day after the last 5th week training; 1-month follow-up; 6-months follow-up On MBEST score; Retained improvements at 1 and 6-month follow-ups; Improved UPDRS. - Chang et al. (2019)[54] 21 (21*,0) *10 PD+FoG; 11 PD-FoG 10 rounds of 50-step SIP movements SIP baseline and cued; (washout interval of at least 1 week between the 2 conditions) Off Cadence; Step variability; Cortical Silent Period; Improved Short Interval Intracortical Inhibition - Chomiak et al. (2017)[50] 11 (11**,0) **6 music; 5 podcast 3 times per week; 10-20 min training sessions Before and after 4-week SIP at-home training On Improvement on dual-task step automaticity with music but not podcast No significant effect of training on FES-I, MoCA, or FOG-Q Byl et al. (2015)[42] 24 (12**,0) **7 EG; 5 CG 6–8 weeks, 12 visits (90 min each) total of 18h of training Mobility, range of motion, balance, and strength exercises baseline; 6-weeks post-training NM Number of falls; Step length; Improved strength, Berg balance and range of motion. Non significant improvements in velocity and TUG time. 34 STUDY TYPES Papers were classified as 2 different types: cross-sectional studies, when evaluating only single session short-term effects of cueing (41 studies) [26]–[41], [43]–[49], [52], [56], [57], [60], [62]–[75]; or longitudinal, when patients trained with cueing strategies for a continuous period, and multisession data analysis were performed in order to evaluate the long-term usage effects of cueing (9 studies) [42], [50], [51], [53]–[55], [58], [59], [61]. Ratanasutiranont et al. [39], although being longitudinal, it does not evaluate the long-term use of cueing strategies in motor rehabilitation, it only divides data collection during 2 sessions 1-week apart. Therefore, this paper outcomes were analysed as a cross-sectional study. PD PATIENTS AND CONTROL GROUPS The total number of PD patients summarized in this review, therefore the sum of all articles PD patients’ samples, is 1543, being 43 the median sample size. Of the 50 articles analyzed, 18 studies [26]–[28], [30], [31], [43]–[46], [48], [49], [54], [56], [66]–[69], [72], [74] separated PD patients in both freezers (with FoG), and non-freezers (without FoG). Other 16 studies considered only freezers for the clinical protocol [32]–[35], [37]–[41], [45], [47], [62]–[64], [70], [73], and 1 study used compared one group of freezers with healthy controls [47]. Therefore, this resulted in a total of summing a total of 642 freezers and 378 non-freezers, with corresponding median sample sizes 20 and 20, respectively, and leaving 523 PD patients indiscriminate regarding suffering or not of FoG. The maximum sample size for PD patients was 154, found in Capato et al. [46] and the minimum was 4, found in Ratanasutiranont et al. [57]. The main factors reviewed to characterize PD patients were age (48 studies) [26]–[49], [52]– [75]; Movement Disorder Society-Unified Parkinson’s Disease Rating Scale (MDS-UPDRS) (43 studies) [26]–[49], [52]–[75]; disease duration [26]–[49], [52]–[75]; and Hoehn and Yahr scale (H&Y) [26]–[49], [52]–[75] (both in 40 studies); gender (34 studies) [26]–[49], [52]–[75]; and New Freezing of Gait Questionnaire (NFOG-Q) or other FoG questionnaires (35 studies) [26]–[29], [31]–[37], [39]–[43], [46], [49]–[54], [56], [57], [60], [62]–[64], [67], [68], [70], [71], [73], [74]. Less than 50% of the studies characterized patients evaluating other clinical scales such as and Montreal Cognitive Assessment scale (MoCA) (15 studies) [26]–[28], [30]–[32], [41], [43], [50], [52], [56], [63], [68], [69], [73]; Mini-Mental State Examination (MMSE) (16 studies) [37], [38], [40], [43], [46]–[49], [54], [55], [57]–[59], [62], [64], [70]; Falls Efficacy Scale (FES–I) (9 studies) [26], [27], [29], [32], [46], [50], [53], [68], [74]; Parkinson’s Disease Questionnaire (PDQ-39) (2 studies)[33], [43]; and Mini Balance Evaluation Systems test (MBEST) (2 studies) [52], [53]. Fewer studies also evaluated less 35 common scales such as Geriatric Depression Scale – 15 (GDS-15), Postural Instability and Gait Disorder (PIGD) Score, Frontal Assessment Battery (FAB) Total Score, Beck’s Depression Inventory (BDI) Score, Royal’s Clock Drawing task (CLOX 1&2), and State-Trait Anxiety Inventory (STAI) Scores. Besides clinical scales, PD patients were characterized according to Levodopa Equivalent Daily Dosage (LEDD) (16 studies) [26], [27], [29], [32], [35], [36], [44], [46], [52], [60], [62]–[64], [66], [67], [73]; height (14 studies) [26]–[28], [34], [36], [43], [44], [49], [55], [59], [64]–[66], [75]; and weight (10 studies) [26]–[28], [34], [36], [49], [55], [59], [65], [75]. In some papers (4 studies) [63] usability questionnaires were also performed to analyze patients feedback about the cueing device. It is important to note that these characteristics evaluations were not considered as clinical parameters analysed, since only provide information to characterize the population being studied. As for medication state, the majority of the studies were assessed on medication (31 studies) [26]–[28], [30]–[32], [34], [35], [37]–[41], [43]–[46], [49]–[51], [53], [55], [57]–[59], [61]–[63], [65], [66], [74]. Alternatively, 12 studies [54], [71], [75], opted for assessing protocols during off medication, and 5 papers [33], [36], [48], [52], [70], tested both on and off states, to exclude bias of dopaminergic medication intake. Byl et al. [42] and Sijobert et al. [68] did not mention in which medication state patients were assessed. Only 11 studies used healthy control groups to compare results with the PD populations being sampled [26]–[29], [31], [44], [46]–[49], [67]. This resulted in a total of 203 healthy controls, being the median sample size for healthy control groups 18. The maximum sample size for controls was found in Delval et al. [48], and the minimum, besides not having control group, was found in Willems et al. [49], with 30 and 9 healthy participants, respectively. It is important to notice that 6 studies used only one group of PD patients to perform different cued and uncued tasks, in random order, to compare the cueing vs. non-cueing effects over mobility [52], [53], [57], [60], [65], [71], [75]; 2 studies distributed PD patients in two alternative groups being one cued and other Placebo, which meant patients used a similar device but without providing any type of cue [55], [58]; 4 other papers divided patients in experimental groups, receiving cueing, and control groups, that did not train with the cueing devices [42], [59], [61]; and 2 studies opted for other alternative sample groups organizations [50], [53]. Chomiak et al. [50] and Capato et al. [53] different group assignments are worth explaining separately. Chomiak et al. [50] divided patients in 2 groups, both receiving auditory stimuli in different manners: one group received music (5 patients) and other received stimuli in a form of a podcast (6 patients). These 2 groups were analysed for understanding the motor and FoG effects of a more rhythmic 36 induced auditory cue (music) when compared to an auditory cue that requires almost dual-task focus (podcast). Alternativelly, Capato et al. [68] divided PD patients in 3 groups: one group that received Rhytmic Auditory Stimuli (RAS)-supported multimodal balance training (56 patients), one group that performed the same training but without cueing (50 patients) and a third control group that instead of performing the training, received an educational program explaining them the RAS training effects in addressing PD motor symptoms. These third group was assigned to compare the training and cueing effects with a group that although not being cue trained, is alerted of the intentions behind cue training. It is [53]also important to note that Byl et al. [42], besides evaluating PD patients also [42]evaluated patients that suffered from strokes, but for the purpose of this research that group of patients was excluded. To summarize the different combinations and organizations of sample groups reviewed there were identified 9 studies that only divided freezers vs. non-freezers [30], [43], [54], [56], [66], [68], [69], [72], [74]; 9 studies that divided freezers, non-freezers and healthy controls [26]–[28], [31], [44], [46], [48], [49], [67]; a single study that compared a freezer group with healthy controls [47]; a single study that compared one PD group and one healthy control group [29]; 16 studies that only used one group of freezers [32]–[35], [37]–[41], [45], [47], [62]–[64], [70], [73]; 6 studies that only used one group of PD patients [52], [57], [60], [65], [71], [75]; 2 studies that analyzed one Placebo group vs. one experimental group [55], [58]; 4 studies designing one experimental vs. one control group of PD patients [42], [51], [59], [61]; and 2 studies opting for two different cued conditions [50], [53]. Ultimately, the mean number of experimental group participants was 19, 10 for Placebo group participants and 13 for control group with PD patients. The respective minimum and maximum numbers were 7 and 22 for experimental group sizes, 10 and 10 for Placebo group sizes; and 5 and 20 for control with PD patients group sizes. INCLUSION AND EXCLUSION CRITERIA Although not listed in the table, it was also important to analyse the most prevalent inclusion and exclusion criteria considered, for further consideration during the protocol design. It is important to note that not all studies presented these criteria. Four studies did not mention any exclusion criteria, but presented inclusion criteria [57], [62], [73], [75]; other four studies did not mention the inclusion criteria, but presented exclusion criteria [30], [44], [67], [72]; and five studies did not mention both [36], [38], [58], [63], [68]. The main and inherently present inclusion criteria considered was participants being diagnosed with PD, were 10 studies inclusively mentioned performing the diagnosis according to the UK brain bank criteria [26], [27], [31], [32], [40], [41], [53], [60], [64], [65]. The great majority of the studies only 37 included patients that were able to walk or stand independently (24 studies) [26], [27], [31], [34], [35], [37], [39], [42], [43], [46], [47], [49]–[51], [53]–[55], [59]–[62], [65], [66], [73], to grant that patients were able to perform the motor tasks; integrated a certain range of stages of H&Y scale (21 studies) [26]– [28], [31], [32], [34], [42], [47], [49], [51], [53]–[57], [61], [65], [69], [71], [74], [75], being the most prevalent including patients that ranged from stage 2 to 3 or patients ranging from stage 1 to 4; were evaluated with history or presence of FoG (19 studies) [32]–[35], [37], [39]–[41], [45], [47], [48], [57], [59], [60], [62], [64], [65], [70], [73], using for that purpose NFOG-Q or UPDRS-III; were able to perceive visual, auditory or vibratory stimuli (12 studies) [26], [27], [31], [35], [37], [46], [53], [54], [62], [70], [73], [75], mainly using Snellen chart to evaluate visual acuity; having a certain age (12 studies) [26]– [28], [31], [32], [35], [41], [46], [54], [64], [74], [75], being mostly common to consider patients that have at least 50 years; and absence of cognitive impairment or not being demented (11 studies) [29], [32], [37], [46], [47], [51], [53], [54], [57], [62], [74], frequently evaluated considering MoCA score or MMSE score larger than 24. As for exclusion criteria, most of them complete the inclusion criteria by reinforcing determined aspects. Participants most predominantly were excluded if presented other neurodegenerative or health disorders (34 studies) [26]–[35], [39]–[44], [46], [48]–[52], [54], [56], [59], [60], [64]–[67], [69]–[72], such as Alzheimer’s disease, stroke or gait disorders; if presented dementia or severe cognitive impairment making them unable to follow instructions (25 studies) [26]–[31], [37], [39], [41], [43], [44], [48]–[50], [54], [55], [59]–[61], [64], [65], [67], [69], [70], [72]; visual, hearing or somatosensory perception impairments (22 studies) [30], [32]–[35], [37], [39], [43], [44], [46], [50]–[52], [55], [59], [60], [64]–[66], [69], [70], [72]; and if patients were unable to walk or walk independently (10 studies) [26]–[28], [41], [45], [52], [56], [64], [69], [71]. It is also worth mentioning that 5 studies also considered to exclude patients that had DBS or other functional brain surgery, to prevent risk of bias [46], [47], [49], [64], [74]. MOTOR TASKS To evaluate patients’ motor performance using different cueing methodologies, it was essential to analyse different motor tasks. Essentially, motor tasks can be divided in 3 categories: walking tasks (34 studies) [27]–[33], [36]–[46], [51]–[53], [55], [57]–[59], [61]–[65], [67], [68], [70], [75]; turning tasks (13 studies) [26], [34], [36], [38], [40], [47], [49], [60], [63], [69]–[71], [74], such as 8-shaped turning task, 180º left U-turn, turning in place and 360º turns; [27]–[33], [36]–[46], [51]–[53], [55], [57]–[59], [61]–[65], [67], [68], [70], [75] gait initiation tasks (10 studies) [27], [28], [37], [40], [48], 38 [56], [66], [68], [72], [73], such as first step initiation, sitting-to-standing or stepping-in-place (SIP), and they can be performed individually or combined. Different walking tasks were performed such as passing through a doorway (7 studies) [30], [37], [38], [44], [45], [63], [70], with the purpose of exposing patients to potential FoG triggering scenarios; range of motion tasks (1 study) [42] and balance training tasks (4 studies) [42], [53], [58], [61], that consist in walking or standing in different surfaces, or during conditions such as eyes closed, to analyse postural and motor control aspects. Timed Up and Go (TUG) task is an example of combining both gait initiation, walking and turning tasks. TUG task consists of evaluating gait performance when standing up from chair, walking, turn around, walk back and sit-down. In this research 7 studies performed TUG tasks [35], [39], [44], [52], [53], [61], [68], being Capato et al . [53] the only study that performed TUG dualtask. Therefore, it is also important to highlight dual-task training (6 studies) [31], [35], [39], [44], [51]–[53], [56], [61], [63], [68]. Dual-task training consisted of simultaneously performing gait and attentional tasks. Three studies performed dual-task training by counting backwards by 7 [63] or by 3 [56], while walking (2 studies) [56], [63] or performing TUG dual-task test [53]. Similarly, Graham et al. [31] performed a forward digit span test while walking. As for Sijobert et al. [68] performed SIP dual-task test, by naming backwards the months of the year while stepping in place. Lastly, Ginis et al. [51] performed walking while reciting as many words from a letter as possible. OUTCOMES All 5 different parameters (Spatiotemporal, Kinetic, Kinematic, Physiological or Clinical) were analysed to evaluate cueing effects over PD patients gait performance and in mitigating motor symptoms, such as FoG. Different result outcomes were obtained, meaning positive (45 studies) [26]–[36], [39]– [44], [46]–[48], [50]–[55], [57]–[75], and negative (24 studies) [28], [30], [37], [38], [41], [42], [44], [45], [47]–[51], [56], [61]–[64], [66], [67], [69], [70], [72], [73], cueing effects over motor response were analysed. Spatiotemporal parameters were the parameters that showed greater improvements with cueing (31 studies) [26]–[29], [31], [32], [34], [36], [39], [42]–[44], [46], [48], [52], [54], [55], [57]–[60], [62], [64]–[68], [72]–[75], mainly increasing patients’ velocity while performing gait tasks [26]–[28], [32], [39], [46], [52], [58], [65], [66], [75], and cadence [44], [46], [54], [59], [67], therefore reducing time to complete tasks [34], [36], [39], [57], [68], [74]; increased step length [32], [42], [43], [62], [66], [73], and stride length [27], [28], [31], [39], [44], [46], [55], [58], [59], [64], [65], [67], [75]; decreased step 39 or stride time [31], [34], [46], [48], [62] , and variability [31], [52], [54]; allowing better task performance and more stable/smoother gait patterns [60], [69]. It is worth noting that in Mancini et al. [69] reducing the number and velocity of turns was not considered a real negative aspect, since it was hypothesized that this decrease was due to patients being more aware and focused during the turning tasks with cueing. Patients’ smoothness during turning increased which indicated that they were more cautious during turning performance, making them slower but in alternative providing them an improved gait pattern and reducing FoG episodes. As for Clinical parameters, the main improved outcome was the number and duration of FoG episodes (14 studies) [30], [33], [34], [36], [39]–[41], [47], [57], [65], [67]–[69], [73]. Geerse et al . [41] initially showed negative effects as the % time FoG and number of episodes increased at first. However, it was considered a habituation period, since as patients proceeded with training those negative effects were overcome. Other clinical positive effects of cueing were reducing number of falls [42]; improved MBEST score [53], although at 4-week follow-up not maintained [51]; improved UPDRS [53], and better balance performance [42], [61]. Tosserams et al. [52] verified also a relationship between some clinical and spatiotemporal results, such as lower MDS-UPDRS III scores being linked to improved gait variability, meaning patients with less severe disease symptoms had better improvements in gait pattern; and higher MoCA scores linked to greater efficacy of external cueing, meaning less cognitive impaired patients showed better improvements when using cues. Another clinical outcomes worth mentioning are Peppe et al. [58] conclusions that showed that patients with higher H&Y score, therefore more severe disease stage, showed better improvements in MDS-UPDRS III, meaning their motor symptoms and disease perception may have improved. The Placebo group also showed some improvements in MDS-UPDRS III, meaning patients perception of the disease may have been also improved by the sense of having a ’false’ cueing device even when it was not delivering cues. As for Kinematic positive effects, the main improvements were improved APA’s duration [28], [48], [72] and amplitude [72]; improved head rotation and head-pelvis separation during turning [47]; improved forward Center of Mass (CoM) velocity and sideways sways [73]; improved lower limbs joints and feet strike heads [31], [59]; and increased knee flexion and ankle dorsiflexion [67]. However, some studies showed that APA’s amplitude did not increase amplitude in all patients, indicating that patients with higher UPDRS III scores decreased APA’s amplitude [28], meaning APA’s improvements may not be pronounced for all patients. 40 In other hand, kinetic outcomes gave information about pressure and forces effecting during different mobility tasks. The main improvements observed were improved positioning [35] and improved Center of Pressure (CoP) sways in the medio lateral plane [61]. Lastly, physiological outcomes mainly served as information to relate with other parameters. The main physiological results obtained were improved saccade, therefore improved rapid movement of the eyes between fixation points, using visual cues [31], meaning patients were more focused on preparing their movement when provided visual feedback; and greater uptake of FF-CT in the better cueing response group [40], which may indicate a potential connection between the compensatory effects of visual cueing and the dopaminergic function in specific brain regions in PD patients. One study indicated improvements of Cortical Silent Period (CSP) and Short Interval Intracortical Inhibition (SICI), meaning that cortical inhibition may have been reduced by auditory cues, leading to better gait performance [54]. As for EMG results, Schlenstedt et al. [56] mentioned that, although first step decreasing when in dual-task performance, contraction of the tensor fasciae latae increased, therefore it was noticed an improvement of muscle activity using cueing. 2.3. DISCUSSION 2.3.1. RQ1: HOW HAVE CUEING STRATEGIES BEEN USED AND APPLIED FOR TARGETING MOTOR SYMPTOMS OF PD? Cueing strategies have been employed to target PD motor symptoms by providing external stimuli to improve movement and gait. They can be divided essentially in 3 types according to the sensory aspect targeted: visual, auditory, or somatosensory cueing; and it was possible to understand their implementations separately or combined. Cues were delivered also via electronic devices, non-eletronic strategies, or both methods combined. The great majority of the studies implemented Visual or Auditory cueing, essentially and respectively via laser devices to project laser lines for patients to step or metronomes to provide rhythm for patients to follow, while Somatosensory cueing was present in less than a third of the reviewed papers. Of this third, more than half combined Somatosensory cues with Visual and/or Auditory stimuli, which means that there are a lack of research evaluating Somatosensory cues alone in addressing motor symptoms such as FoG. The main somatosensory devices used were vibratory systems, such as VibroGait, to provide vibratory cues at a certain rhythm or when detecting a determined gait-event or phase. 47 3. SOLUTION OVERVIEW This dissertation is the progress of several clinical studies that aim to analyse and compare different cueing strategies, to conclude which may cause better motor performance improvements in a cross-sectional study; develop a novel strategy; and practice a long-term protocol, in PD patients, using +sensBand and evaluating the motor effects of those strategies in gait rehabilitation, focusing on FoG and QoL. To achieve this, it is necessary to first consider the materials necessary to develop the novel solution. In this chapter, it is outlined: (i) an overview of the system focusing on the main hardware components, (ii) an introduction and description about the different strategies used and (iii) a description of the MATLAB App implemented for further data processing. 3.1. HARDWARE In which concerns the technologies used in this project, it is important to take into account two different systems: the +sensBand system, responsible for providing both somatosensory cueing, visual cueing and for data collection (Chapter 4, 5 and 6), and the Xsens motion tracking system with Awinda, used for gait assessment validation in Chapter 6. Fig. 7 and Table 3 summarize the technologies used. Table 3. Technologies used during this dissertation Technologies Purpose Data collected Parameters evaluated +sensBand Provide different control strategies of somatosensory and visual cueing, during patients gait performance, and Collects acceleration and angular velocity. Allow to determine gait velocity, cadence, step length, step time, stride length, stride b) Xsens a) +sensBand Fig. 7. Technology systems used. 48 simultaneously collecting gait metrics. time, asymmetry, and variability Xsens Validate +sensBand data collection and provide additional biomechanical outcomes Collects motion data of each body segment position First, and since it is the main component to address in this project, it is important to start by understanding the +sensBand system and the different components integrated. The +sensBand device itself includes 6 systems: a processing system; an actuation system, that includes vibration motors responsible for somatosensory cueing and 2 laser projectors for visual cueing; a sensory system, including 4 IMU’s responsible for data collection and a camera for door detection; a memory system; a communication system, responsible for communicating via Bluetooth with a smartphone App, designed for device configuration, and via SSH with the clinical desktop App, designed for data processing; and a power supply, that consists of a Charmast powerbank 10000. All 5 systems are located inside a 3D printed box, except for the sensory system, whose IMU’s are located throughout the waistband textile itself, at 4 different locations around the waist: back; left side; right side and navel point. +sensBand allows to monitor axel accelerations and angular velocities on all these locations, simultaneously or not, to assess gait metrics, while the actuation system provides visual or vibrotactile cues according to the control strategy selected. The Xsens system contains 6 wireless Motion Trackers (MTw), that consist of miniaturized IMU’s that incorporate 3D accelerometers, gyroscopes, magnetometers and a barometer, therefore providing 3D orientations as well as linear acceleration, angular velocity, magnetic field and pressure data. This entails more detailed metrics than the +sensBand sensory system, allowing to validate the +sensBand collected data (acting as a ground truth) but also providing more information to assess patients motion outcomes. The IMU’s are positioned on the patients’ body (e.g. lower and upper legs, upper arms, wrists, and pelvis), using Velcro body straps. To charge and calibrate the MTW’s, an Awinda Station is connected to an outlet, via DC power connector, and to a computer, or using an Awinda USB Dongle, via USB, allowing synchronization of data assessment when using the different MTW’s simultaneously. 3.2. CONTROL AND MONITORING STRATEGIES The +sensBand originally offered five possible control strategies, that can be selected using the smartphone App, including (i) gait-event driven cueing; (ii) context awareness cueing; (iii) turning 49 phase-dependent cueing; (iv) postural stability phase-dependent cueing and (v) continuous cueing. Based on the literature review conclusions, it was further developed a novel sensory cueing strategy, the (vi) Fog prevention cueing. For this project, only strategies (i), (ii) and (vi) were explored, therefore only those will be further discussed. Besides providing different cueing strategies, the +sensband also provides a motion monitoring option. All the different strategies can be selected to configure the +sensband device using a smartphone app, as shown in Fig. 8. In case the user pretends to not use any type of cueing, but rather only monitor the patient’s gait, it should select the Motion monitoring option directly. When the Sensory cueing option is selected instead, a page with the different cueing strategies available will appear and the strategy pretended should be selected and then configured. Each of the strategies will be further explained in this section. Fig. 8. +sensBand smartphone App. 3.2.1. MOTION MONITORING PD motor symptoms are commonly present during patients' everyday tasks. Changes in patients gait pattern such as reduced amplitude of movement, slowness, and shorter steps, can be an indicator for more severe symptoms such as FoG. Therefore it is of enormous importance to analyze patients gait metrics. By these means, the +sensBand was developed considering also the development of a motion monitoring strategy. This strategy activates only the IMU’s of the +sensBand, to collect acceleration and angular velocity data, for further analysis of gait metrics. Gait event detection requires the Back IMU data as a reference, so it is important to guarantee that the IMU is placed correctly at the lower back (approximately above L5). The +sensBand integrates a 50 gait event detection algorithm that receives the IMU data in real-time, then calibrates it to eliminate measurement errors, applies motion compensation to adjust for trunk rotations, and filters data to detect significant changes indicative of gait events.Then, the sensory data runs trough a real-time adaptitive algorythm, that allows to identify five key gait events: Heel strike (HS), the initial contact (IC) moment; Foot flat (FF); Mid stance (MSt); Heel off (HO); and Toe off or the final contact (FC) of the foot. This algorithm was developed specifically for PD patients, containing adaptative thresholds for each gait event, that are updated along the motion monitoring assessment. The adaptative thresholds are recalculated and updated periodically in sets of 10 gait cycles, considering the data of the previous gait cycle set. The motion monitoring strategy is also automatically activated when using all the different cueing strategies, since it requires to activate motion monitoring to obtain the patient gait data file collected. 3.2.2. GAIT EVENT-DRIVEN CUEING Several cueing strategies of +sensBand were explored, to comprehend each functionality but also to select which should be adapted and implemented in the clinical protocols. Gait event-driven cueing consists in providing closed-loop vibratory stimulation, triggered every time a right-foot toe-off is detected. Toe-off gait event was selected since literature reports observation of inadequate swing phase before the occurrence of FoG episodes [79]. The vibratory frequency and time on which the cueing strategies are provided, which motors will apply the vibration, and the frequency on which the data will be collected are all selected/configured using the smartphone App for +sensBand. During the use of this strategy all motors were selected to provide vibrations around the waist, and were set at 200Hz to vibrate during 200ms, according to [19]. Fig. 9. Gait event-driven cueing delivery schematic. The gait event detection algorithm plays an important role also in this cueing strategy. By segmenting the gait events, it is possible to detect the TO’s of both feet, that are then distinguished considering that the right foot HS’s presents angular velocity in the Antero Posterior plane ( gyr_AP ) below 0, and left HS’s present gyr_AP above 0 [80]. Closed-loop vibration (200ms at right TO) 51 When the TO’s are detected and they correspond to the right foot TO’s, the motors are activated, and the patient receives the vibrations set at the respective duration and frequency. 3.2.3. CONTINUOUS CUEING Literature review as shown that one of the main strategies for temporal orientation of PD patients was the use of auditory cues set as metronomes, since it provides a fixed rhythm for patients to follow and try to match their pace. However, auditory cueing comprises several disadvantages concerning practical aspects, considering that audio can be intrusive. Somatosensory cueing, on the other hand, when applied at a fixed rhythm can be also used with the same purpose but in a more specific and discrete manner. Continuous cueing, like a metronome, consists in providing continuous vibratory open loop cueing for a defined period of time and at a certain frequency. This strategy was configurated in order to provide vibrations during 2s pulses, set at 200 Hz, according to [19]. Fig. 10 shows the schematic of this cueing mode delivery. Fig. 10. Continuous feedback delivery schematic. 3.2.4. FOG PREVENTION CUEING By analysing the literature review it was noted that, overall, the cueing strategies that predominantly resulted in higher gait improvements and in higher FoG episodes reduction were the visual cues. When applied at open loop providing patients spatial orientation on where to place each step, such as tape lines, acted as active strategies for patients to overcome FoG. Somatosensory cues also had great potential for PD motor rehabilitation and in preventing FoG episodes, especially when applied at a closed loop rhythm, allowing to be personalized for each patient's gait performance to help them be more aware on when to step. Combining both open loop visual cues with closed loop somatosensory cues, therefore, may have greater potential in targeting FoG, since it can act as both an active and preventive strategy, allowing a more complete sensory stimulation of the patients. Therefore, it was developed a novel sensory cueing Open-loop vibration (2s pulses) 52 strategy using the +sensBand system that simultaneously provided both two cueing modalities, the FoG prevention cueing. FoG prevention cueing initially provides vibration at a fixed continuous rate (open-loop) to help gait initiation, along with the projection of two fixed laser lines in front of the patient (also open-loop). After starting walking, the laser lines remain fixed and open-loop. However, the vibrotactile cues change to closed-loop after the first TO of the right foot being detected. Fig. 11 represents the cueing strategy delivery. Fig. 11. FoG prevention cueing delivery schematic. Closed-loop vibration is enabled at the right foot’s toe-off gait event during 200 ms and is set at 200 Hz, using the same algorithm for gait event detection as the gait event-driven strategy. The vibration is set to 2s pulses. Since the start of walking usually elicits FoG, rhythmic open-loop vibration is enabled before gait initiation, reinforcing the first step execution [56], [73]. Therefore, after detecting the first toeoff, the control strategy switches to closed-loop vibration to personalize walking according to the users' self-selected pace. Visual cues are continuously provided by activating the lasers to project two lines in front of the participants. These lines move along with users so they are always seeing the next two steps where the foot should be placed. Therefore, visual cues are applied to assist during a FoG episode, while vibrotactile cues also assist during gait initiation but during walking forward, turning, and passing through doors are applied also to prevent the occurrence of FoG episodes. 3.3. DESKTOP APP (MATLAB) In order to further process and analyse the collected gait data, and calculate spatiotemporal gait metrics, it was developed a desktop clinical App using MatLAB (Fig. 12). Open-loop vibration (2s pulses) Closed-loop vibration (200ms at right TO) Open-loop laser lines (adjusted to participants step length) 53 Fig. 12. Desktop clinical App. This App was designed to read the IMU data files obtained using the +sensBand, process the data using several filters, and then running the data through a finite state machine (FSM) to detect IC’s (HS) and FC’s (TO) for each foot, to further calculate gait metrics. The FSM algorithm was developed to emulate the FSM of the +sensband itself. However, to avoid the accumulation of any errors, such as leg distinction errors, the Matlab FSM includes two additional conditions that ensure that after detecting a gait cycle of the right leg, the next gait cycle must be detected for the left leg, and vice versa. Instead of 5 states, the improved FSM contained a total of 10 states, corresponding to the different gait events in order but for both right and left feet (e.g. HS_right, FF_right, TO_left, MSt_right, HO_right, HS_left, FF_left, TO_right, MSt_left and HO_left). If the FSM detects consecutive gait cycles on the same leg, the state resets to correct any potential errors. At TO it was inserted the gyr_AP condition for leg distinction. After detecting right or left HS, the subsequent events (FF, MSt, and HO) are associated to that same leg, except for the TO event that is associated to the opposite leg, in line with the descriptions of [81], [82]. Since it is important to ensure that after a right leg detection the following metrics should be of the opposite leg, and vice versa, the algorythm was set by first detecting the 5 events for the right leg gait cycle ( HS_right, FF_right, TO_left, MSt_right and HO_right), and then following with the same sequence of gait events but for the left leg. If the events were detected with the gyr_AP condition for leg distinction detecting a right or left leg, and the following and previous legs were detected as the opposite legs, the gait events were correctly detected, therefore were inserted in the IC and FCs arrays for gait metrics calculation. Else, it meant that two gait cycles for the same leg were detected consecutively, presenting an error. In this case, the respective gait events detected in that gait cycle were removed, and Nans were added on those positions instead. This approach flags errors, with Nans marking cycles where the expected leg alternation failed, thus preventing inaccurate gait metrics from being calculated. 54 Depending on which foot the patient initiates gait, two FSM cycles were created: one starting with the left foot first and other starting with the right. For better understanding of the improved FSM algorithm, Fig. 13 presents a flowchart with the order of gait event detection highlighted by the color corresponding to the first foot detected in the FSM algorithm. When using the FSM_Right_first cycle, events are detected in the order indicated by the blue numbers, whereas in the FSM_Left_first cycle, events follow the order marked by the red numbers. Ultimately, the input acceleration and angular velocities components (𝑎𝑐𝑐𝑛,𝑔𝑦𝑟𝑛), are processed through the FSM algorithm to identify the respective IC’s and FC’s for each foot (𝐼𝐶𝑅,𝐹𝐶𝑅,𝐼𝐶𝐿,𝐹𝐶𝐿). Fig. 13. Flowchart of the improved FSM for gait segmentation MatLab algorithm. 55 After processing and detecting ICs and FCs of each foot, gait metrics were calculated for both right and left legs, including step time and length; stride time and length; stance, swing and double support phase; gait speed; cadence and number of steps taken. Additionally, asymmetries and standard deviations were also computed by comparing metrics of both feet and analysing variability, respectively, for a more comprehensive understanding of the patient's gait characteristics. Table 4 presents the equations used to obtain each spatiotemporal gait metric calculated. Table 4. Methods for estimating gait spatiotemporal metrics represented by domain, metric, formula and measurement units Domain Metric Formula Units Pace Step length 𝑚𝑒𝑑𝑖𝑎𝑛((2√𝐿ℎ−ℎ2,ℎ=∬ 𝑎𝑐𝑐𝑉 𝐼𝐶2𝑖 𝐼𝐶1𝑖),(2√𝐿ℎ−ℎ2,ℎ=∬ 𝑎𝑐𝑐𝑉 𝐼𝐶1𝑖+1 𝐼𝐶2𝑖)) Meters (m) Velocity 𝑚𝑒𝑑𝑖𝑎𝑛(𝑆𝑡𝑒𝑝 𝑙𝑒𝑛𝑔𝑡ℎ1𝑖 𝑆𝑡𝑒𝑝 𝑡𝑖𝑚𝑒1𝑖 ,𝑆𝑡𝑒𝑝 𝑙𝑒𝑛𝑔𝑡ℎ2𝑖 𝑆𝑡𝑒𝑝 𝑡𝑖𝑚𝑒2𝑖) Meters per second (m/s) Cadence 𝑚𝑒𝑑𝑖𝑎𝑛(𝑉𝑒𝑙𝑜𝑐𝑖𝑡𝑦1𝑖∙60 𝑆𝑡𝑒𝑝 𝑙𝑒𝑛𝑔𝑡ℎ1𝑖,𝑉𝑒𝑙𝑜𝑐𝑖𝑡𝑦2𝑖∙60 𝑆𝑡𝑒𝑝 𝑙𝑒𝑛𝑔𝑡ℎ2𝑖) Steps per minute (step/min) Rhythm Step time 𝑚𝑒𝑑𝑖𝑎𝑛((𝐼𝐶2𝑖−𝐼𝐶1𝑖),(𝐼𝐶1𝑖+1−𝐼𝐶2𝑖)) Seconds (s) Stance phase 𝑚𝑒𝑑𝑖𝑎𝑛((𝐹𝐶2𝑖−𝐼𝐶1𝑖),(𝐹𝐶1𝑖+1−𝐼𝐶2𝑖)) Percentage (%) Swing phase 𝑚𝑒𝑑𝑖𝑎𝑛((100−𝑠𝑡𝑎𝑛𝑐𝑒𝑝ℎ𝑎𝑠𝑒1),(100−𝑠𝑡𝑎𝑛𝑐𝑒𝑝ℎ𝑎𝑠𝑒2)) Percentage (%) Variability SD step length 𝑆𝐷(𝑠𝑡𝑒𝑝𝑙𝑒𝑛𝑔𝑡ℎ1,𝑠𝑡𝑒𝑝𝑙𝑒𝑛𝑔𝑡ℎ2) Meters (m) SD velocity 𝑆𝐷(𝑣𝑒𝑙𝑜𝑐𝑖𝑡𝑦1,𝑣𝑒𝑙𝑜𝑐𝑖𝑡𝑦2) Meters per second (m/s) SD step time 𝑆𝐷(𝑠𝑡𝑒𝑝𝑡𝑖𝑚𝑒1,𝑠𝑡𝑒𝑝𝑡𝑖𝑚𝑒2) Seconds (s) Assymetry AS step length |𝑠𝑡𝑒𝑝𝑙𝑒𝑛𝑔𝑡ℎ1−𝑠𝑡𝑒𝑝𝑙𝑒𝑛𝑔𝑡ℎ2| Meters (m) AS velocity |𝑣𝑒𝑙𝑜𝑐𝑖𝑡𝑦1−𝑣𝑒𝑙𝑜𝑐𝑖𝑡𝑦2| Meters per second (m/s) AS step time |𝑠𝑡𝑒𝑝𝑡𝑖𝑚𝑒1−𝑠𝑡𝑒𝑝𝑡𝑖𝑚𝑒2| Seconds (s) 1 and 2: represent the first and second leg according to the order detected (right leg first or left leg first); i: instant; L: distance from floor to where the back IMU is placed in the user body; h: refers to the double integration of vertical acceleration; 𝒂𝒄𝒄𝑽: refers to the vertical acceleration; SD: standard deviation; AS: asymmetry 3.4. CONCLUSIONS Following an in-depth analysis of the current cueing strategies and technologies for gait rehabilitation in Parkinson's disease (PD) patients, it was found that the integration of advanced systems like the +sensBand and Xsens holds significant potential. These technologies not only provide diverse cueing modalities but also offer robust data collection and validation capabilities, which are essential for evaluating motor effects, particularly in addressing FoG and improving quality of life. Thus, this dissertation focuses on exploring and implementing these technologies within the framework of ongoing clinical studies aimed at developing and assessing novel cueing strategies for PD. To this end, the dissertation provides a detailed overview of the system components, the control and 56 monitoring strategies employed, and the tools used for data processing and analysis. This work is part of the broader +sense project, contributing specifically to the advancement of gait rehabilitation techniques using state-of-the-art equipment such as the +sensBand system and the Xsens motion tracking system. 63 I think the various functions of the system are very well integrated. 4.43±0.83 I thought there was too much inconsistency in this system. 1.48±0.92 I would imagine that most people would learn to use this system very quickly. 4.38±1.09 I found the system very cumbersome to use. 1.29±0.73 I felt very confident using the system. 4.34±0.76 I needed to learn a lot of things before I could get going with this system. 1.45±0.90 Likert scale: 1 – strongly disagree; 2 – disagree; 3 – indifferent; 4 – agree; 5 – strongly agree. 4.2.1. GAIT METRICS Table 9 shows that pace was the domain that presented more significative differences. In this gait domain, step length increased the most in SG2 (≈18.95%). Velocity only showed significant differences between the CG and each study group (SG1 or SG2), increasing with both cueing strategies in comparison to the CG (≈10.70% for SG1 and ≈20.14% for SG2). No significant differences were noticed for velocity in between the two study groups (SG1 and SG2). As for cadence, only were noticed significant differences between CG and SG1, where SG1 showed a greater increase of this metric (≈8.05%). Both study groups were able to counteract the prototypical pace behaviour of PD, improving pace metrics with cueing. Similarly, the rhythm domain presented significant differences for the swing phase between the CG and each study group (SG1 or SG2). Both groups that used the two cueing modalities showed greater improvements of swing phase (≈5.36% for SG1 and ≈8.96% for SG2), by presenting higher swing phases. No significant differences were observed for step time or stance phase. By considering the swing phase improvements, it is possible to understand that both cueing strategies were able to counteract the prototypical rhythm behaviour of PD. As for the variability and asymmetry aspects of gait, different results were observed. For the variability domain, significant differences were noticed for SD step length and SD velocity. For both metrics it was noticed that cueing strategies worsened gait variability by following the prototypical behaviour of PD. However, in SD step length, SG1 presented significant differences when compared with SG2 indicating that gait event-driven cueing (SG1) had better results for SD step length than continuous cueing (SG2) (≈46.79% improvement between SG1 in comparison to SG2). No significant differences were noticed between CG and SG1. SD velocity, on the other hand, only presented significant differences between CG and SG2. Those differences resulted in worsened SD velocity for the SG2 (≈45.94%). Similar results were noticed for AS step length and AS step time, where SG2 performed worse than CG (results worsened by almost twice the CG value). Therefore, it is possible to note that continuous cueing worsened 64 both variability and asymmetry aspects of gait performance. Gait event-driven cueing, when compared to the continuous cueing, provides less variability for step length during gait performance. Overall, the CG presented less dispersion of the SD values, followed by SG1 and lastly SG2. SG2 was the group that presented higher dispersion of the SDs. 4.2.2. SYSTEM USABILITY SCALE (SUS) Considering each patient’s total SUS score, it was obtained a mean±SD score of 86.32±6.90%. This result indicates that the +sensBand was perceived with an excellent usability overall, being evaluated as an easy-to-use (4.77±0.46% for question 3), cohesive (4.43±0.83% for question 5), intuitive (4.38±1.09 for question 7), and comfortable system (1.29±0.73 for question 8). Overall, patients also felt safer and more confident about their gait performance when using the system (4.34±0.76% for question 9). However, some more neutral scores were noticed when answering the first question about if patients would like to use the system frequently (3.77±1.20%). A good part of the patients was still capable of walking autonomously, without any major gait changes, therefore when asked this question, did not feel the need to use a system daily to improve their gait. Some comments were also noted regarding the design of the device not being the most appealing for day-to-day use, especially to use in public. The weight of the system was also commented about to be further improved, since it may not be the most comfortable for daily usage. On another hand, some patients also noted that although at first impact feeling a bit intimidated by the bulky design, they quickly adapted to it and accepted it. During the trials, it was not reported any discomfort while using the device. 4.3. DISCUSSION Considering the prototypical motor behaviour of PD, this section intends to analyze and discuss the gait metrics results to understand if PD patients improved or not their gait performance. It is important to compare also both cued groups' results to provide insights about which of the two cueing methods may be more beneficial for PD patients' motor rehabilitation, and therefore achieve the main goal of this study. The results obtained showed that overall, both cueing strategies were able to improve patients' gait performance by targeting the pace and rhythm aspects of gait, therefore mitigating the hypokinetic and bradykinetic gait patterns of PD. When comparing both cued groups with the control group it was possible to note improvements of step length (≈18.95% for the SG2), velocity (≈10.7% for the SG1 and 65 ≈20.14% for the SG2), cadence (≈8.05% for the SG1), and swing time (≈5.36% for the SG1 and ≈8.96% for the SG2). Although both gait event-driven (SG1) and continuous (SG2) vibrotactile cueing presented improvements for both pace and rhythm, continuous feedback provided greater improvements overall (≈19.54% in mean for pace and ≈8.96% for rhythm) than gait event-driven cueing (≈9.38% in mean for pace and ≈5.36% for rhythm), that only showed greater improvements for cadence. However, when comparing directly SG1 and SG2 results, mixed conclusions were taken depending on the metrics being analyzed. Only step length and SD step length metrics presented significant differences between the two study groups (SG1 and SG2), corresponding to two different aspects of gait (pace and variability, respectively). For step length, the group with continuous feedback (SG2) had greater improvements than the one receiving gait event-driven cueing (SG1). On the other hand, when considering the SD step length, gait event-driven cueing showed improvements when compared to the group receiving continuous feedback. Moreover, the SG2 group when compared to the CG group did not present any improvements of gait variability and asymmetry metrics, instead it was noticed an increase, therefore worsen, of SD step length (≈60%), SD velocity (≈45.94%), AS step length (≈114.95%) and AS step time (≈165.57%) with the use of open loop vibration. Considering these results, it was hypothesized that continuous feedback, by delivering the vibrations at 2s pulses similarly to a metronome, caused more abrupt improvements in pace and rhythm than closed-loop vibration and also higher dispersion of SDs. Although SG2 presenting higher velocity and step length than the other groups, the resultant cadence is lower than the SG1. SG2 patients took approximately 3.41 steps per each 2s passed, meaning that completed approximately one gait cycle of one foot (from IC to the next IC of the same foot) during each 2s receiving the vibrations. This suggests that patients may unintentionally followed the rhythm of the vibrations and tried to fill the pulse duration by taking less steps but longer and slightly quicker. However, since these are more abrupt changes of pace than the changes noticed in the SG1, it caused higher instability of gait translated by worsened variability and asymmetry. Gait event-driven cueing, on the other hand, by providing the vibrations at a rhythm personalized to the patient’s own pace, it consists of a more subtle strategy. Therefore, more subtle changes of the gait pattern were noticed, but less gait variability is noticed since patients perform an improved version of their own gait instead of forcing their gait to match the cues. This highlights the potential of gait event-driven cueing which should not be discarded since it causes great improvements of pace and rhythm still and may surge as a better choice of strategy in preserving patients' gait variability. Also, since the study groups consisted in three independent populations the results may be affected by patients different clinical characteristics such as H&Y stage 66 and UPDRS III that was more advanced in SG1 when compared to both CG and SG2 (SG1 showed mean and SD of H&Y of 1.86±0.76 points, indicating unilateral and axial towards bilateral involvement, and UPDRS III score of 19.44±12.43 points, corresponding to a mild mobility stage). Also, although the CG group has more PD patients with FoG (9 freezers) than SG1 and SG2 (5 freezers each), the SG1 freezers showed higher scores of the NFOG-Q (21.40±7.47). These clinical characteristics may indicate that patients receiving the gait event-driven cueing presented more advanced stages of the disease, therefore presenting worse mobility and gait patterns. Ultimately, patient variability may have affected the results of the study, and providing a cross-over analysis following the protocol in the future should be considered to improve the study conclusions. Lastly, the SUS results of the usability aspects of the device have shown that overall, the +sensBand system was evaluated with an excellent usability score, being acceptable for further implementation in longer trials. However, in the future, the design should be improved to become more appealing for daily usage and less heavy to improve comfort for more extended periods of use. 4.4. CONCLUSIONS Both open and closed-loop cueing strategies, as expected, improve gait performance by mitigating the hypokinetic and bradykinetic aspects of gait. Although overall continuous feedback causing greater improvements of pace and rhythm than gait event-driven cueing, it increased gait variability and asymmetry. Considering the different advantages of both cueing strategies, they both have great acceptability and potential for motor rehabilitation of PD patients, moreover if explored in a combined new strategy. 67 5. USABILITY STUDY This study serves to introduce the FoG prevention cueing, with healthy participants for a primary validation of the usability and workload of this novel strategy, while simultaneously using the +sensBand device in different environments, to evaluate also the potentiality of being further used as a daily solution. Therefore, the primary objective of this study is to validate the usability and workload of this new cueing strategy in healthy participants as a preliminary step before testing in the PD population. By evaluating the system's impact on gait temporal metrics and usability through the System Usability Scale, this research aims to ensure the device's effectiveness and comfort in daily use. Positive findings in this initial phase will pave the way for further trials focused on PD patients, ultimately aiming to improve their mobility and overall quality of life. 5.1. METHODOLOGY 5.1.1. PARTICIPANTS A total of fifteen healthy participants (5 males and 10 females; 30.47 ± 14.09 years; higher education level; daily use of technologies) were included in this study. To be eligible to integrate this study, participants had to meet the following criteria: (i) not present any neurodegenerative disorders and gait impairments; (ii) be able to walk independently without any assistance; and (iii) sign informed consent to participate in the study. The study was approved by the ethical Ethics Committee CEICVS 006/2020, according to the Declaration of Helsinki and Oviedo Convention. 5.1.2. PROTOCOL After instrumenting participants with the +sensBand waistband (Fig. 15 a)) and configurating the FoG prevention feedback strategy, a first calibration trial was executed, instructing them to take one step forward to adjust the +sensBand lasers to match the user’s self-selected step length. After calibrating the lasers, three walking trials were performed, in one of three possible paths (in work or home-based scenarios). Fig. 15 b) shows the different environments used. a) Instrumented subject 68 Fig. 15. FoG prevention feedback usability study test (a)) and courses (b)). The work-based environment was a longer walking path (total of 64m) performed on a hallway of University of Minho, at Azurém Campus, and only 3 participants were selected for this scenario due to their availability to travel to the site and familiarization with the course. As for the home-based scenarios (A and B), they took place in two different houses, common to the participants. Considering space limitations of the different housings, shorter walking courses were considered (A: total of 13m; B: total of 16m). The purpose of at home trials was to assess participants in a secure and familiar setting, common to their daily routine, while using the device. By these means, participants were allocated to each course considering their day-to-day familiarization with environments A or B. At the start of each trial, participants were instructed to stand in place for 20s, while receiving the open-loop vibrations and fixed laser lines. After those 20s, participants started walking and the vibrotactile mode changed to closed loop after detecting the first right foot TO. During the walking path, participants turned 90 degrees, passed through at least one door, turned 180 degrees and walked back to the start position. During each trial the researchers accompanied the participants to confirm that the +sensBand was providing the cues correctly. 5.1.3. DATA ACQUISITION Each participant's demographic information was collected considering age, gender, education level, and technology usage. To assess system's usability and workload, the System Usability Scale (SUS) and NASA Task Load Index (TLX) questionnaires were answered anonymously to reduce risk of bias. SUS questionnaire allowed to evaluate user experience aspects such as willingness to use the system often, complexity, ease of use, consistency, comfort and convenience. As for the NASA TLX questionnaire, three types of demands were evaluated (mental, physical, and temporal), but also performance, effort, and frustration felt. The A B 3m 3.50m 5m 3m 20m 7m 5m b) Study walking courses 69 statements from the SUS and NASA TLX questionnaires were rated using a Likert scale, ranging from 1 (Strongly disagree) to 5 (Strongly agree) and 1 (Low) to 21 (High), respectively. Resulting SUS scores ranged from 0 to 100, allowing to be categorized as awful, poor, reasonable, good, or excellent for scores lower then 51, from 51 to 68, equal to 68, from 68 to 80.3, or greater then 80.3, respectively. Similarly, NASA TLX resulting scores ranged also from 0 to 100, categorizing workload as very high, high, somewhat high, medium, or light for scores higher then 80, from 50 to 80, from 30 to 50, from 10 to 30, or lower then 10, respectively. 5.1.4. DATA PROCESSING Gait metrics were also assessed by the +sensBand during each trial, by activating the motion monitoring option while performing the FoG prevention feedback. The data files obtained were then processed using the desktop clinical App, to obtain the following temporal gait metrics: stride, step, stance, and swing times, and cadence. Transitions from HS to FF, and from FF to TO were identified and stored in lists to detect Initial Contacts (IC) and Final Contacts (FC) during gait. Those lists were then used to calculate each temporal parameter as follows: step time and stride time were calculated as the temporal distance from IC of one foot to the following, and one IC to the next of the same foot, respectively; stance time was calculated as the temporal distance from IC to FC of the same foot; swing time was calculated as the difference between stride and stance time (respective phases were obtained by converting those times in percentages); and cadence was calculated by counting the number of steps, corresponding to the length of the list of IC, and dividing it by the gait duration, therefore time distance between first and last IC. The mean temporal gait metrics were calculated for each participant considering the gait cycles performed during the protocol. Finally, mean and standard deviation of the temporal gait metrics obtained from all participants were presented for the results section. 5.2. RESULTS Table 11 presents both demographics and results obtained for each participant based on the questionnaire answers evaluating the usability and workload of the FoG prevention feedback using the +sensBand. Furthermore, mean and standard deviations (mean±SD) of both SUS and NASA TLX scales are presented. Table 12 then presents the mean results of participants' temporal gait metrics. 70 Table 11. Healthy participants’ age, gender (F – female; M – male), Education level, Technology usage, Walking course path (A – Home-based path A; B – Home-based path B; W – Work-based), SUS and NASA TLX scores, and corresponding qualitative levels (colored through green to red from better through worst usability, respectively) ID Age (years) Gender Walking path SUS (score, level) NASA TLX (score, level) P1 19 F B 82.50 Excellent 18 Medium P2 M B 80 Good 11 Medium P3 22 F W 90 Excellent 9 Light P4 A 75 Good 30 Somewhat high P5 A 100 Excellent 18 Medium P6 A 92.50 Excellent 11 Medium P7 24 F A 90 Excellent 15 Medium P8 25 F A 82.50 Excellent 15 Medium P9 27 F W 92.50 Excellent 21 Medium P10 F W 52.50 Poor 19 Medium P11 M B 77.50 Good 37 Somewhat high P12 30 M B 72.50 Good 21 Medium P13 55 M B 82.50 Excellent 8 Light P14 57 F B 97.50 Excellent 10 Medium P15 59 M B 77.50 Good 29 Medium (mean±SD) 30.47±14.09 F B 83±14.08 Excellent 18.1±9.36 Medium Table 12. Healthy participants’ mean step, stride, stance and swing times, stance and swing phases, and cadence, and respective total means and standard deviations ID Step time (s) Stride time (s) Stance time (s) Stance phase (%) Swing time (s) Swing phase (%) Cadence (steps/min) P1 0.69 1.37 0.91 66.20 0.46 33.80 87.20 P2 0.68 1.34 0.91 68.20 0.43 31.80 88.24 P3 0.80 1.41 1.02 72.24 0.39 27.75 74.69 P4 0.63 1.25 0.77 61.66 0.48 38.33 95.15 P5 4.11 6.29 4.68 74.40 1.61 25.60 14.58 71 P6 1.10 1.95 1.44 73.46 0.52 26.54 56.31 P7 2.58 3.92 2.78 71.59 1.15 28.41 29.35 P8 1.14 1.94 1.29 66.52 0.65 33.48 54.27 P9 0.59 1.11 0.82 73.78 0.29 26.22 102.41 P10 1.05 2.06 1.33 64.40 0.74 35.60 56.85 P11 0.89 1.63 1.11 67.98 0.52 32.02 70.34 P12 1.78 3.16 2.24 70.60 0.92 29.40 33.61 P13 1.23 2.13 1.47 69.83 0.66 30.17 49.99 P14 2.48 4.63 2.83 61.72 1.80 38.28 25.19 P15 2.75 4.62 2.95 64.72 1.67 35.28 25.41 (mean±SD) 1.50±0.99 2.59±1.52 1.77±1.07 68.49±4.09 0.82±0.49 31.51±4.09 57.57±27.17 5.2.1. SYSTEM USABILITY SCALE (SUS) SUS scores were predominantly excellent, with nine participants (8 females and 1 male) scoring between 82.50 and 100, aged 19-57 years, and predominantly training in home-based scenarios. Five participants (1 female and 4 males) scored good usability (72.50-80), with ages ranging from 19 to 59 years, and training mainly in home-based walking path B. Ultimately, one participant (female) scored poor usability (52.50), with 27 years and walking in work-based walking path. Overall, the results obtained were positive, considering an excellent average score (83 ± 14.08). 5.2.2. TASK LOAD INDEX (NASA TLX) Only two participants (1 female and 1 male) scored the lowest workload level (light), ranging from 8 to 9, aged 22 and 55, and executing work and home-based B walking paths, respectively. Medium workload was the predominant score level, with eleven participants (8 females and 3 males) scoring between 10 and 29, aged 19-59 years, and predominantly training in both home-based courses. Two participants (1 female and 1 male) presented fewer positive results, scoring somewhat high (30 and 37), with ages 22 and 27 years, and training in home-based paths A and B, respectively. Although not reaching the best workload score level, the mental demand was still positive, with an overall medium average level (18.10 ± 9.36). 5.2.3. TEMPORAL GAIT METRICS By processing the data metrics obtained, it was possible to calculate the mean and standard deviations considering all participants for step time (1.50s±0.99s), stride time (2.59s ± 1.52s), stance time (1.77s ± 1.07s), swing time (0.82s ± 0.49s) and cadence (57.57steps/min ± 27.17steps/min). 72 Although not reaching the ideal stance and swing phases (60% and 40%, respectively) the means obtained were approximate of the expected resulting in 68.49% ± 4.09% and 31.51% ± 4.09%, respectively [23]. 5.3. DISCUSSION Regarding usability, participants generally evaluated the cueing strategy and +sensBand with an excellent SUS score, therefore considering it as highly usable [20]. Ultimately, the system was wellreceived, considering it intuitive to use and user-friendly among the participants. However, it was reported one poor level of SUS score. Although answering anonymously, most of the participants answered the questionnaires in the presence of the researchers so that they could be able to clarify any doubts that may arise. However, some participants couldn’t answer right after the trials the questionnaires, answering further an online version without the researcher’s presence. Ultimately, those cases reported difficulty in interpreting the questionnaire. Questions evaluating the necessity/likability of including this device in their daily routines were answered negatively in this case due to the participant focusing on their lack of need for the device rather than its usability. Also, questions about the performance and technological integration aspects of the device were not answered correctly, since some patients opted to give a neutral score since they didn’t intend to comment on those aspects. All these misinterpretation errors ultimately may have affected the final score. In the future, clearer instructions should be considered to clarify the scales, as long as prioritizing answering the questionnaire immediately after the motor performance, with the researchers still present. As for workload perception including both physical and mental demands, analysis of NASA TLX scores revealed varying levels of workload experienced by participants during the task. While most reported a medium workload level, some experienced lighter or somewhat higher demands. It is important to note that one participant, during the walking course reported that the laser projected lines, since were advancing as the participant also progressed, made the participant frustrated during the first trial trying to always step over them. This may have translated into a high score of workload perceived by this participant. However, the rest of the participants did not report any type of frustration and considered the laser lines an intuitive cueing modality, even more than the vibrations. When comparing home with workbased walking paths, the results of perceived workload tend to be positive in both environments. Homebased environments may present low mental demand due to familiarity and comfort. However, the two 79 episodes occurring off the trials but during the sessions, i.e. during rest intervals or other moments, were not counted as FoG observations since were not present when using the strategy but were noted about the tasks done when occurring. During the evaluation time points established in the protocol, NFOG-Q answers were also collected to provide a clinical evaluation of FoG evolution before and after intervention, and if the results were maintained after a week (follow-up). Clinical data was collected, evaluating H&Y, MDS-UPDRS III and PDQ-39 clinical scales at specific time points of the protocol (pre-training, post-training and follow-up), before the motion monitoring trials during each protocol time-point, to evaluate PD clinical evolution before and after receiving cues. Similarly, some questionnaires were answered to evaluate different parameters of the cueing strategy used along the sessions. NASA TLX answers were collected at the end of each training session, to evaluate cognitive and physical demand of the strategy, as well as answering a personalized questionnaire about patients’ perception of the different cues received during the trials, to evaluate how the cueing strategies were being used along the training sessions. The personalized questionnaire contained four questions, answered on a range from 1 to 5, as shown in Appendix B. At the end of the training protocol (after completing the session 12 cued training trials) patients answers of the SUS were also collected to provide end-users evaluation of the usability of the cueing strategy after long-term usage. During pre-training, both Xsens and +sensBand allowed to acquire, at 100Hz, and collect acceleration, angular velocity and gait events data. The +sensBand also allowed to collect these metrics in the middle of training (session 6 or 7), at post-training and follow-up. These two systems were used at pre-training to compare both metrics obtained at baseline and therefore more thoroughly validate the +sensBand system performance for gait event detection. The +sensBand system was also used during the different time points to understand patients' gait performance evolution before and after using the cueing strategy. 6.1.4. DATA PROCESSING Due to the reduced number of patients, for a better understanding of the results data processing consisted in performing descriptive analysis with the acquired data with the results varying along the respective protocol timeline. At first, it was important to analyze patients calendarized sessions and transform that calendar in a timeline. Considering patients' availability and different starts of the protocol, the 12 sessions of training were spaced differently for each patient. Since the pre-training and the first session, which occurred on the same day, were the reference sessions starting at day 0, each session that followed was marked on a timeline by considering the number of days that passed since this reference. Ultimately, two protocol 80 timelines were obtained, a timeline ranging from 0 to 54 days for patient P1, and another ranging from 0 to 46 days for patient P2. Overall, the protocol for patient P1 took 54 days to complete, where training took 47 days, and for patient P2 took 46 days to complete, where training took 39 days. According to each type of acquired data and respective time points for data acquisition, the timeline of the respective graphic varied according to the sessions in between which the data were collected. This means that, data that was collected continuously along each session, such as number of FoG episodes perceived in-between sessions and number of FoG episodes observed in each session during trials, was graphically presented showing the results along each session of the full protocol (from pre-training to follow-up). Data that instead of being collected at each session was collected only at the protocol evaluation time points (pre-training, mid-training, post-training and follow-up), such as +sensBand motion monitoring resulting gait metrics, Xsens gait metrics, NFOG-Q results, MDS-UPDRS III results and PDQ-39 results, was graphically presented showing each result along the respective evaluation timepoint session. Lastly, data evaluated during the training sessions (+sensBand gait metrics acquired while receiving cues, NASA TLX results, and cue perception questionnaire results) was displayed graphically along a timeline ranging from the first training session to the last (session after pre-training to session before post-training). It is important to note that H&Y, since it is a more specific scale designed to assess the overall stage of PD progression, was only evaluated before and after receiving the cued training (pre-training, post-training, and follow-up). This decision was made as the scale is not as sensitive to short-term changes throughout the intervention, unlike the other metrics. Therefore, given its relatively stable nature over time, the H&Y results were better suited to be displayed in a table format. SUS results were also opted to be displayed in table format since were only assessed once, after completing the training sessions and therefore were not evaluated along a timeline. Lastly, the +sensBand gait data was processed, similarly to Chapters 4 and 5, to obtain the following gait metrics: step length, velocity and cadence, for the pace domain; step time, stance phase and swing phase, for the rhythm domain; SD of step time, velocity and step time, for the variability domain; and AS of step time, velocity and step time, for the asymmetry domain of gait. The Xsens data had to be imported from MVN, selecting acceleration, angular velocity but also foot contact to serve as ground truth for the gait event detection. Then the files had to be converted to mat type files, to be able to be read in Matlab. Lastly, using the Clinical Desktop App the Xsens foot contact data was processed obtaining the same gait metrics as the +sensBand gait data. Since it was 81 assessed three trials in each timepoint, it was calculated at each evaluation time points the mean and SD of the trials for each patient +sensBand resulting gait metrics. Similarly, at pre-training the mean and SD of the trials for each patient were calculated for both +sensBand and Xsens gait metrics. The +sensBand results, since are important to be analyzed along the different assessment moments, were then displayed graphically along the evaluation timepoints timeline. 6.2. RESULTS For better comprehension of the results, this section is segmented in four subsections according to the different types of data acquired: (i) FoG results, where observed episodes during trials, perceived episodes in between sessions, episodes duration evolution along the protocol timeline and NFOG-Q results were considered; (ii) clinical scales results, including H&Y, UPDRS III and PDQ-39, evaluated only at protocol time-points; (iii) training questionnaires results, including questionnaires done continuously in each training session, such as NASA TLX and a custom-made perceived cues questionnaire, but also at the end of training such as SUS questionnaire; and (iv) gait metrics results, analyzing both Xsens and +sensBand gait collected data during the evaluation time-points but also continuously along the training sessions. 6.2.1. FOG Fig. 18, Fig. 19 and Fig. 20 present the graphics obtained for the FoG results. Fig. 18. Graphics of FoG perception in between sessions vs. FoG occurrences during trials, for patients P1 and P2. When analysing the results of patients' perceived FoG in between sessions vs. FoG occurrences during trials (Fig. 18), it is noticeable that, for both patients, the perceived curve (black line points) tends to have unpredictable behaviour, but the observational curve (red line points) tends to decrease. 82 Patient P1 did not freeze during any trial along the protocol timeline. As for FoG episodes reported in between sessions, patient P1 perceived a maximum of 5 episodes in between sessions 5 and 6, that took place 4 days apart. When analysing only the evaluation time points, from pre-training to post-training the number of FoG episodes perceived increased (≈50%), and from post-training to follow-up assessment the number of FoG episodes perceived improved greatly by decreasing to 0 episodes occurring in between post-training and follow-up assessments. Patient P2 presented more cohesive observational results being noticeable a decrease of FoG during trials along the three main evaluation time points (≈100% in between pre-training and post-training, and ≈66.67% in between pre-training and follow-up). From post-training to follow-up, however, it was observed an increase (≈33.33%) of the number of FoG episodes occurring after 7 days without receiving cues. Fig. 19. Graphics of gait task responsible for FoG (Gait initiation, Turning, or Passing through doorways) along the sessions timeline and respective total percentage, both for patient P2. It is important, to analyse also which tasks lead to freezing (Fig. 19). In pre-training the main FoG-inducing task was turning during the walking course (≈100% of FoG at session 0). Overall, counting all sessions, turning was the main FoG-inducing task (≈53% of all FoG episodes), followed by passing through doorways (≈30% of all FoG episodes), and lastly gait initiation (≈17% of all FoG episodes). However, at follow-up, the only FoG episode noticed occurred during gait initiation at the start of the first monitoring trial. The maximum number of FoG episodes observed during trials occurred in session 3, consisting of a total of 19 FoG episodes (≈5.26% during gait initiation, ≈63.16% during turning, and ≈31.58% while passing through doors). As for P2 perception of FoG in-between sessions, the maximum number of FoG episodes reported were 17 episodes at follow-up, after 7 days without receiving cueing. However, as far as considering the ratio between number of episodes reported and number of days passed, the maximum number of episodes between sessions was reported in between session 5 and 6 (≈4 ratio), reporting 16 FoG episodes along 4 days without cueing. 83 Fig. 20. Graphics of NFOG-Q scores along each time-point evaluation, for patients P1 and P2. When analysing the NFOG-Q answers along the protocol time-points (Fig. 20), overall patient P2 reported higher scores than patient P1. Patient P1 along the protocol time points presented increasingly worse results of NFOG-Q, presenting a maximum score of 10 points at both post-training and follow-up. As for patient P2, the results varied along the time points, being noticed worst score results at midtraining, with 19 points. At both pre-training and post-training P2 obtained the same minimum NFOG-Q score of 15 points, however at follow-up the score increased by one point. 6.2.2. CLINICAL SCALES Table 14, Fig. 21 and Fig. 22, show the table and graphics obtained for clinical scales results. Table 14. H&Y score evolution before and after using the FoG prevention cueing, for patients P1 and P2 Stage Description Impairment Pre-training Posttraining Followup 0 Without any signs of the disease. 1 Unilateral impariment. Mild to moderate 1.5 Axial and unilateral impairment. P1 P1 2 Bilateral disease without balance deficit. P1 P2 P2 2.5 Mild bilateral disease, with recovery in the 'push test'. P2 3 Mild to moderate bilateral disease; some postural instability; ability to live independently. 4 Severe disability, still able to walk or stand without assistance. Severe 84 5 Confined to bed or a wheelchair unless assisted. When considering the H&Y results (Table 14) it is possible to observe that for both patients, when comparing the moments before (pre-training) and after (post-training) receiving the cued training with the FoG prevention strategy, both patients improved their motor impairments. These score improvements were maintained at follow-up. Both patients showed an improvement of 0.5 score points, however, P2 presented still higher H&Y stage and, therefore worse PD symptom severity when compared to P1. Both MDS-UPDRS III and PDQ-39 graphics (Fig. 21 and Fig. 22, respectively) showed an overall similar profile, with a tendency to take a U-shape format. Fig. 21. UPDRS III scores evolution along each evaluation timepoint, for patients P1 and P2. Fig. 22. PDQ-39 scores evolution along each evaluation timepoint, for patients P1 and P2. When observing the MDS-UPDRS III results (Fig. 21) overall both patients decreased their scores, meaning improvements of motor function at each evaluation time point when compared to pre-training (≈50%, ≈35.71% and ≈21.43% for P1, and ≈52.17%, ≈82.61% and ≈78.26% for P2, at mid-training, posttraining and follow-up, respectively). However, both patients experienced a slight regression towards the end of the follow-up period (≈22.22% for P1 and ≈25% for P2, regression from post-training). Patient P1 showed maximum improvements of UPDRS III scores at mid-training, while patient P2 showed maximum 85 improvements at post-training. Patient P2, although noting a regression at follow-up, presented a better MDS-UPDRS III score at follow-up than at mid-training, explaining the less prominent U-shape pattern of this graphic. Patient P1, although showing results at follow-up worse than mid-training, those results were still better than pre-training. As for the PDQ-39 score results (Fig. 22) overall both patients also experienced an improvement in QoL over the evaluation time points. When comparing both patients, P1 showed greater variations along each timepoint’s score, while P2 showed a less prominent improvement, with a more stable pattern overall. P1 achieved a minimum score, meaning maximum improvement felt of QoL, at mid-training (≈16.10% improvement from pre-training). At post-training, although having a small regression from the mid-training results (≈5.74% regression), still showed some degree of improvement when compared with the pre-training PDQ-39 score (≈11.29% improvement). A small regression towards the end of the followup period ocurred, but without returning to pre-training levels. Patient P2 however showed maximum improvement at post-training (≈19.45% improvement from pre-training), improving gradually from pretraining to mid-training (≈15.28% improvement) and continuing to improve from mid-training to posttraining (≈4.92% improvement) reaching its peak at this timepoint. At follow-up, for both patients although being noticed a regression of the results since post-training (≈7.30% regression for P1 and ≈8.64% regression for P2), QoL improved since pre-training (≈4.81% improvement for P1 and ≈12.49% improvement for P2). Overall, patient P2 although showing a more stable pattern, presented higher improvements of QoL from pre-training to both post-training and follow-up. 6.2.3. TRAINING QUESTIONNAIRES Fig. 23, Fig. 24 and Table 15 present the graphics and table of the training questionnaires subsection. Fig. 23. NASA TLX scores evolution along each training session, for patients P1 and P2. 86 When analysing the mental and physical workload (Fig. 23) at the last training session both patients P1 and P2 reported lower NASA TLX scores, therefore improving along the sessions the mental effort felt during the cued training (improvements of ≈37.5% for patient P1 and ≈67.74% for P2). Patient P2 showed overall better results of NASA TLX scores along the training timeline, however both patients started the protocol considering the workload at a somewhat high level (39 score for P1 and 31 score for P2) but decreased it to a medium workload at both mid and post-training (28 and 25 scores, respectively, for patient P1, and 10 scores at both moments for P2). Fig. 24. Cue perception Questionnaire answers for each question (Q1: “How perceptible was the laser?”, Q2: “With what frequency did you use the laser cues?”, Q3: “How perceptible was the vibration during the trial?”, and Q4: “How perceptible was the vibration in the first 20s of the trial?”) evolution along each training session, for patients P1 and P2. Cue perception results (Fig. 24) showed some variability across the two patients. Both patients reported high perception of vibration in the first 20 seconds of trials (Q4), however during the walking trials, the vibrations perceived at right toe-offs were lightly perceived (Q3). Overall, both patients could perceive the laser very well (Q1), however, only used it slightly (Q2). Patient P1 did not felt the need to use the laser and felt slightly confused by it. Patient P2 on the other hand reported using the laser lines right before FoG inducing tasks, to prevent from freezing. When the number of FoG episodes during the sessions started to diminish, this patient also started to report the need of using the laser less. Table 15. SUS answers for each individual question, for patients P1 and P2 SUS Questions P1 P2 I think that I would like to use this system frequently. 4 4 I found the system unnecessarily complex. 1 2 I thought the system was easy to use. 5 5 I think that I would need the support of a technical person to be able to use this system. 1 1 I think the various functions of the system are very well integrated. 4 3 87 I thought there was too much inconsistency in this system. 1 3 I would imagine that most people would learn to use this system very quickly. 5 5 I found the system very cumbersome to use. 1 1 I felt very confident using the system. 4 5 I needed to learn a lot of things before I could get going with this system. 1 1 Likert scale: 1 – strongly disagree; 2 – disagree; 3 – indifferent; 4 – agree; 5 – strongly agree. As for the SUS results (Table 15) both patients showed overall exellent usability reaching total scores of 92.5 and 85, for patients P1 and P2, respectively. 6.2.4. GAIT METRICS First, the metrics obtained at pre-training (baseline) with the +sensBand were compared with the ones captured by the Xsens that served as our ground truth. If the +sensBand resulting metrics fell within the range of the respective Xsens metrics were deemed acceptable, therefore validating those +sensBand results. Table 16 summarizes the resulting metrics obtained using both devices (Xsens and +sensBand) for each patient P1 and P2. Table 16. Gait metrics obtained with the Xsens vs. +sensBand, for each patient P1 and P2 at pre-training Xsens +sensBand Gait metrics P1 P2 P1 P2 (pre-training) Mean±SD Mean±SD Mean±SD Mean±SD Step length [m] 0.565±0.021 0.484±0.035 0.515±0.015 0.604±0.047 Velocity [m/s] 0.974±0.009 0.988±0.114 0.950±0.075 1.062±0.089 Cadence [steps/min] 102.572±0.877 115.035±10.295 112.549±10.931 89.864±8.606 Step time [s] 0.585±0.005 0.513±0.032 0.550±0.050 0.597±0.030 Stance phase [%] 75.420±16.937 57.974±8.206 78.909±0.701 61.693±4.425 Swing phase [%] 24.580±16.937 47.050±8.206 21.091±0.701 38.307±4.425 SD Step length [m] 0.140±0.027 0.058±0.036 0.283±0.037 0.426±0.113 SD Velocity [m/s] 0.234±0.022 0.227±0.026 0.479±0.073 0.489±0.104 SD Step time [s] 0.901±0.147 0.113±0.062 0.228±0.155 0.137±0.022 AS Step length [m] 0.059±0.032 0.087±0.061 0.242±0.207 0.576±0.130 AS Velocity [m/s] 0.128±0.122 0.347±0.010 0.216±0.177 0.133±0.083 AS Step time [s] 0.038±0.012 0.148±0.107 0.167±0.110 0.177±0.051 Bold highlights the +sensBand metrics validated when compared with the Xsens metrics. 88 Overall, velocity, stance phase and swing phase were the only metrics validated for both patients. Patient P1, however, presented both cadence, AS step length and AS velocity within the respective XSens metrics range, therefore those +sensBand metrics were also validated for this patient. As for patient P2, only SD and AS of step time were further validated. Lastly, Fig. 25, Fig. 26, Fig. 27, Fig. 28 and Fig. 29, present the graphics of the gait metrics results. When analysing the +sensBand gait metrics graphics along the protocol evaluation timepoints, it is possible to note different patterns along some metrics but also along each individual patient. Fig. 25. Mean step length, respective SD and AS evolution along protocol evaluation timepoints, for each patient. Fig. 26. Mean velocity, respective SD and AS evolution along protocol evaluation timepoints, for each patient. Step length (Fig. 25) and velocity (Fig. 26) were the metrics that showed similar profiles for both patients, improving gradually as expected along the training sessions and reaching the max improvement peak at post-training (≈17.13% and ≈10.85% for patient P1, and ≈20.39% and ≈22.89% for patient P2, improvements of step length and velocity, respectively, from pre-training to follow-up). At follow-up, the results showed a slight regression for both patients and both metrics, however, were still maintained improvements when compared to the pre-training assessment. 95 although being extensive was very well received, being perceived with excellent usability and medium workload. Patients’ motor performance was also improved at post-training, especially when considering improvements of pace (step length and velocity) but also rhythm (step time), reinforcing improvements of the hypokinetic and bradykinetic aspects of gait. 96 7. CONCLUSIONS This chapter provides final reflections and research findings, highlighting the work contribution to knowledge. An overview of the results considered within each chapter is presented, and research questions are answered. Future steps to be considered are also discussed, enhancing the impact and further potential of this project research. This dissertation relied on the +sensBand system project, aiming to improve its strategies, developing and validating a new one for mitigating FoG, and testing its long-term use potential for motor and QoL improvements. The work developed innovated by overcoming some of the previous challenges of the +sensBand research project, concluding research on the efficiency of gait eventdriven cueing and continuous feedback in PD patients, providing a novel strategy suitable for more advanced stages of PD and performing its validation in a longitudinal protocol to allow insights on a rehabilitative perspective of the +sensBand system. This approach was based on gathering and analyzing, both patients’ feedback and motor data, with the aim to enhance patients' motor performance and reduce motor symptoms of PD during daily activities, therefore improving their QoL. The research efforts, along with the encouraging results achieved through experimental methodologies with end-users, played a crucial role in achieving the dissertation goal, as outlined below. The gathered knowledge in Chapter 2, was the motor for the project development, allowing an understanding of the different available cueing solutions being applied in PD patients, as well as how were being applied and their respective effects (KPI1). It was noticed that cueing strategies, particularly somatosensory and visual strategies, show promising results in reducing PD motor symptoms such as FoG. Closed-loop patient-tailored approaches highlighted their potential to significantly improve motor task performance. The literature review also allowed to understand that research should prioritize spatiotemporal outcomes, such as velocity and FoG occurrences, while incorporating regular, consistent training protocols and assessments to ensure sustained benefits with long-term cued training. Technological systems, along with the +sensBand solution components, considering the new implemented and developed strategy, were introduced and described in Chapter 3: (1) the hardware, considering both +sensBand and Xsens systems; (2) the cueing and monitoring +sensBand strategies used; and (3) the clinical desktop App improved for gait data processing. The +sensBand allows with a single device to provide several cueing strategies, through laser line projection and/or vibrotactile motors, while simultaneously collecting patients gait data through an IMU- 97 based system. The Xsens system, by using high-tech IMU-based gait assessments, acts as a ground truth to validate the +sensBand gait assessments. The +sensBand configuration is established via a smartphone app, which allows to select the respective strategy pretended. If it’s intended to only monitor patients gait without providing cueing, the motion monitoring strategy allows to activate only the IMU sensors, collecting patients' angular velocity and acceleration data, and enabling real-time gait event detection. As for the cueing strategies, three modalities were considered. The gait event-driven cueing strategy consists on providing closed-loop vibratory cueing at right-foot toe-offs. The continuous feedback, alternatively, provides open-loop vibratory cueing, set to 2s on-and-off pulses. By considering that combining somatosensory closed-loop cueing and visual open-loop cueing may target FoG and enhance gait performance by instructing when and where to place the foot, respectively, the FoG prevention cueing was developed. Similarly to the gait event-driven cueing, it allows to provide vibrations at right-foot toe-offs, while simultaneously providing open-loop laser line projection set to patients' step length. Since gait initiation was identified as one of the main FoG-inducing tasks, the strategy was adapted to provide open-loop vibration as the continuous feedback strategy until the first right foot toe-off is detected. After detecting this event it automatically switches to the closed-loop mode. It was also presented the clinical desktop app, developed in Matlab, that allows to transform the acquired data into the respective gait metrics being analyzed. The acceleration and angular velocity data run through calibration, filtering and gait segmentation processes (Finite State Machine algorithm) to obtain a list of ICs and FCs for each foot. By using these IC/FCs, gait metrics are calculated for both pace and rhythm domains, but also variability and asymmetry metrics. The following three chapters comprise each of the three studies performed along this dissertation, regarding the respective protocol established, results obtained and conclusions. Chapter 4 presented the on-going cross-sectional study executed in which was accomplished both the on-field familiarization with the +sensBand technologies but also the sample size goal to complete the study (KPI2). This study also allowed to understand the efficacy of both gait event-driven cueing and continuous feedback strategies in improving motor performance, being both capable of improving both pace and rhythm aspects of gait. Overall, there is potential of using both strategies combined, reinforcing the need of development of the FoG prevention cueing strategy. Chapter 5 followed with the execution of a usability and workload study, performed in healthy individuals, for validation of the FoG prevention cueing strategy in different daily walking scenarios. From 98 this study, it was concluded that the strategy was perceived with excellent usability and low workload, therefore granting that it should not overstimulate patients (KPI3). After the validation of the FoG prevention cueing strategy in healthy participants (Chapter 5), Chapter 6 followed with the design and execution of the longitudinal study to address the long-term rehabilitative potential of the +sensBand directly with the end-users. The results, although showing some differences due to different patient characteristics, proved the potential of the FoG prevention cueing in mitigating FoG during training (KPI4). Its effectiveness, however, depends on consistent use. Improvements in motor performance and quality of life were also noted, as long as confirming the excellent usability of the strategy and mid-workload perceived (KPI5). Future work should consider targeting a larger number of FoG patients to participate in the study, to allow greater impact of spatiotemporal results. Along with this dissertation document, the work developed resulted in 2 papers, a journal paper and a conference paper (KPI6). In summary, the results obtained along this dissertation experimental assessments confirmed the effectiveness of the +sensBand system usage in improving PD patients motor performance, reducing motor symptoms and ultimately reaching better QoL, at both short and long-term uses. It was already possible to put the +sensBand to test in home-based scenarios and also at a longitudinal protocol setting. The findings obtained should serve as motivation to explore new approaches such as home-based longterm usage. Some adjustments mainly of design should still be considered. However, overall, this dissertation research project made a significant and grateful contribution to enhancing the daily lives of PD patients through the application of biomedical engineering solutions. 7.1. RESEARCH QUESTIONS The resulting dissertation work presented allowed to answer the RQs outlined in Chapter 1. The concluding answers are presented in summary as followed. RQ1: How have cueing strategies been used and applied for targeting motor symptoms of PD? This RQ was answered in Chapter 2, by revising the literature available about different cueing applications designed to target several motor symptoms of PD. Cueing strategies applied to PD patients primarily involve external stimuli such as visual, auditory, and somatosensory cues, delivered through electronic and non-electronic methods. Visual cues, like laser lines, are mainly used to improve spatial 99 awareness, targeting mainly the pace aspects of gait (improve hypokinetic impairments), while auditory cues, like metronomes, enhance rhythm during gait (improve bradykinetic impairments). Somatosensory cues, often vibratory, are less researched and mostly used in combination with other stimuli. Cueing can be open-loop, providing the cue at a fixed rhythm, or closed-loop, synchronized with patients’ gait performance for a more tailored approach. Studies have focused on using cueing during several walking tasks, often incorporating FoG triggers, to evaluate cueing efficacy using established clinical scales and collecting gait metrics. RQ2: Which is the effectiveness of cueing strategies concerning motor effects and QoL? Chapters 2 and 4, through the literature review executed and the cross-sectional study developed with PD patients testing different cueing strategies, allowed to understand their efficacy in improving both motor symptoms and QoL. Considering the diverse range of cueing options, cueing strategies have shown varied effectiveness in addressing motor symptoms and improving QoL in PD. Short-term studies (cross-sectional studies) indicated that visual and auditory cues significantly improve gait parameters, targeting mainly pace aspects such as cadence, velocity, and step length, and can also reduce the number FoG episodes in more advanced stages of the disease. Somatosensory cues, on another hand, also show potential, but their effectiveness is less consistent. Combining different cue types, for example, visual and somatosensory cues, can enhance both motor and QoL outcomes but may require attention in which concerns avoiding overstimulation. Long-term studies (longitudinal studies) suggest that while initial improvements in motor function and QoL occur, these effects may diminish over time without a continuous use of cueing. Closed-loop cueing appeared to provide better long-term benefits than open-loop methods, mainly in targeting FoG, improving pace (step length and velocity) and some aspects of rhythm (stance phase). Moreover, when comparing the use of open and closed-loop somatosensory cueing strategies in a cross-sectional study executed, it was proven that both cueing delivery methods are efficient in improving gait performance, however different aspects may be improved slightly differently with each method. Open loop cueing although showing greater overall improvements in pace, contributed to higher gait variability and asymmetry. In contrast, closed-loop cueing led to more subtle, stable improvements by aligning with the patient’s natural rhythm, potentially preserving gait variability. RQ3: How does +sensBand cueing system long-term use affect Parkinson’s motor symptoms and QoL? The final study chapter, Chapter 6, by addressing a longitudinal protocol in PD patients, allowed to answer this question considering the results obtained. The promising findings support the potential of 100 +sensBand cueing strategies, in the FoG prevention cueing form, in targeting more advanced symptoms of PD, such as FoG (mitigating FoG ocurrences totally during training and reducing up to 66.67% after a week without receiving cued training) , improving mainly hypokinetic aspects of gait (≈17.82%)., and improving overall perceived QoL (≈15.37%). However, the effectiveness for FoG mitigation depends also on the regularity of training using the FoG prevention cueing strategy, since the majority of the improvements were not maintained at follow-up, meaning that the strategy caused mainly immediate improvements during training that were not maintained at long-term. The improvements observed, especially at post-training, demonstrate that sensory cueing can be a valuable tool for mitigating FoG, but its application must be continuous. These findings, however, are not sufficient to guarantee the use of this sensory solution as a rehabilitative device instead of an assistive device for motor rehabilitation, therefore needing more evidence to conclude about these aspects. 7.2. FUTURE DIRECTIONS Some limitations noticed along this dissertation work noticed consisted of: (i) high variability of PD patients, that may affect group and overall results; (ii) design limitations, regarding the bulkiness of the device; and (iii) small sample of PD patients available to participate in longitudinal studies. These aspects should not discard, however, the potential of further investigation with the device. Furthermore, as patients often reported not perceiving the vibrotactile cues during walking, future work should explore alternative +sensBand cueing strategies, such as subliminal cues, to improve PD motor symptoms, as these may be more seamlessly integrated into patients' daily routine. Lastly, the algorithm for gait-event detection proved to also have some limitations regarding patients gait variability due to the unpredictable nature of the disease. Therefore, to improve real-time gait-event detection and gait metrics calculation performance, the implementation of a machine learning algorithm able to adapt to different gait patterns should be considered instead of the developed FSM algorithm for IC/FC segmentation. 101 8. REFERENCES [1] C. A. Davie, “A review of Parkinson’s disease”, doi: 10.1093/bmb/ldn013. [2] B. Thomas and M. Flint Beal, “Parkinson’s disease”, doi: 10.1093/hmg/ddm159. [3] M. Delgado‐Alvarado, M. Marano, A. Santurtún, A. Urtiaga‐Gallano, D. Tordesillas‐Gutierrez, and J. Infante, “Nonpharmacological, Nonsurgical Treatments for Freezing of Gait in Parkinson’s Disease: A Systematic Review,” Movement Disorders , vol. 35, no. 2, pp. 204–214, Feb. 2020, doi: 10.1002/mds.27913. [4] D. Bega, P. Gonzalez-Latapi, C. Zadikoff, and T. Simuni, “A Review of the Clinical Evidence for Complementary and Alternative Therapies in Parkinson’s Disease,” Curr Treat Options Neurol , vol. 16, no. 10, p. 314, Oct. 2014, doi: 10.1007/s11940-014-0314-5. [5] A. Nieuwboer, “Cueing for freezing of gait in patients with Parkinson’s disease: A rehabilitation perspective,” Movement Disorders , vol. 23, no. SUPPL. 2. 2008. doi: 10.1002/mds.21978. [6] L. Wang et al. , “Effects of Rhythmic Auditory Stimulation on Gait and Motor Function in Parkinson’s Disease: A Systematic Review and Meta-Analysis of Clinical Randomized Controlled Studies,” Front Neurol , vol. 13, 2022, doi: 10.3389/fneur.2022.818559. [7] H. R. Gonçalves, A. M. Rodrigues, and C. P. Santos, “Vibrotactile biofeedback devices in Parkinson’s disease: a narrative review”, doi: 10.1007/s11517-021-02365-3/Published. [8] S. G. Reich and J. M. Savitt, “Parkinson’s Disease,” Medical Clinics of North America , vol. 103, no. 2, pp. 337–350, Mar. 2019, doi: 10.1016/j.mcna.2018.10.014. [9] S. Von Campenhausen et al. , “Prevalence and incidence of Parkinson’s disease in Europe,” European Neuropsychopharmacology , vol. 15, no. 4, pp. 473–490, 2005, doi: 10.1016/j.euroneuro.2005.04.007. [10] N. Giladi and A. Nieuwboer, “Understanding and treating freezing of gait in parkinsonism, proposed working definition, and setting the stage,” Movement Disorders , vol. 23, no. S2, pp. S423–S425, Jul. 2008, doi: 10.1002/mds.21927. [11] A. Nieuwboer and N. Giladi, “Characterizing freezing of gait in Parkinson’s disease: Models of an episodic phenomenon,” Movement Disorders , vol. 28, no. 11. pp. 1509–1519, Sep. 15, 2013. doi: 10.1002/mds.25683. [12] J. Vandenbossche et al. , “Freezing of gait in Parkinson’s disease: disturbances in automaticity and control,” Front Hum Neurosci , vol. 6, 2013, doi: 10.3389/fnhum.2012.00356. 102 [13] A. Schrag, M. Jahanshahi, and N. Quinn, “How does Parkinson’s disease affect quality of life? A comparison with quality of life in the general population,” Movement Disorders , vol. 15, no. 6, pp. 1112–1118, Nov. 2000, doi: 10.1002/1531-8257(200011)15:6<1112::AIDMDS1008>3.0.CO;2-A. [14] L. M. Ambani and M. H. Van Woert, “Start Hesitation — A Side Effect of Long-Term Levodopa Therapy,” New England Journal of Medicine , vol. 288, no. 21, pp. 1113–1115, May 1973, doi: 10.1056/NEJM197305242882108. [15] H. Khalil and J. A. Bajwa, “Barriers and Facilitators in Physical Rehabilitation for Parkinson’s Disease in the Arabian World,” Mov Disord Clin Pract , vol. 2, no. 3, pp. 227–229, Sep. 2015, doi: 10.1002/mdc3.12200. [16] T. C. Rubinstein, N. Giladi, and J. M. Hausdorff, “The power of cueing to circumvent dopamine deficits: A review of physical therapy treatment of gait disturbances in Parkinson’s disease,” Movement Disorders , vol. 17, no. 6, pp. 1148–1160, Nov. 2002, doi: 10.1002/mds.10259. [17] P. Ginis, E. Nackaerts, A. Nieuwboer, and E. Heremans, “Cueing for people with Parkinson’s disease with freezing of gait: A narrative review of the state-of-the-art and novel perspectives,” Ann Phys Rehabil Med , vol. 61, no. 6, pp. 407–413, Nov. 2018, doi: 10.1016/J.REHAB.2017.08.002. [18] C. Cosentino et al. , “One cue does not fit all: A systematic review with meta-analysis of the effectiveness of cueing on freezing of gait in Parkinson’s disease,” Neurosci Biobehav Rev , vol. 150, p. 105189, Jul. 2023, doi: 10.1016/J.NEUBIOREV.2023.105189. [19] H. Gonçalves, R. Moreira, A. Rodrigues, and C. Santos, “Finding Parameters around the Abdomen for a Vibrotactile System: Healthy and Patients with Parkinson’s Disease,” J Med Syst , vol. 42, no. 11, p. 232, Nov. 2018, doi: 10.1007/s10916-018-1087-2. [20] J. R. Lewis, “Item Benchmarks for the System Usability Scale,” 2018. [Online]. Available: https://www.researchgate.net/publication/330225055 [21] N. Muthukrishnan, J. J. Abbas, H. A. Shill, and N. Krishnamurthi, “Cueing Paradigms to Improve Gait and Posture in Parkinson’s Disease: A Narrative Review,” Sensors , vol. 19, no. 24, p. 5468, Dec. 2019, doi: 10.3390/s19245468. [22] K. Horváth et al. , “Changes in Quality of Life in Parkinson’s Disease: How Large Must They Be to Be Relevant?,” Neuroepidemiology , vol. 48, no. 1–2, pp. 1–8, 2017, doi: 10.1159/000455863. 103 [23] U. M. Fietzek, S. J. Schulz, K. Ziegler, and A. O. Ceballos-Baumann, “The Minimal Clinically Relevant Change of the FOG Score,” J Parkinsons Dis , vol. 10, no. 1, pp. 325–332, Jan. 2020, doi: 10.3233/JPD-191783. [24] M. Pimenta, C. Pinheiro, and C. Santos, “Robotic sensory cueing during gait: a usability study with scenarios of daily living,” in 2024 IEEE International Conference on Autonomous Robot Systems and Competitions (ICARSC) , May 2024, pp. 165–170. doi: 10.1109/ICARSC61747.2024.10535942. [25] T. Zhang, D. Meng, D. Lyu, and B. Fang, “The Efficacy of Wearable Cueing Devices on Gait and Motor Function in Parkinson Disease: A Systematic Review and Meta-analysis of Randomized Controlled Trials,” Arch Phys Med Rehabil , Aug. 2023, doi: 10.1016/j.apmr.2023.07.007. [26] J. Das et al. , “Visual Cues for Turning in Parkinson’s Disease,” Sensors , vol. 22, no. 18, 2022, doi: 10.3390/s22186746. [27] S. Stuart, J. Wagner, S. Makeig, and M. Mancini, “Brain Activity Response to Visual Cues for Gait Impairment in Parkinson’s Disease: An EEG Study,” Neurorehabil Neural Repair , vol. 35, no. 11, pp. 996–1009, Nov. 2021, doi: 10.1177/15459683211041317. [28] Y. Russo, S. Stuart, C. Silva-Batista, B. Brumbach, G. Vannozzi, and M. Mancini, “Does visual cueing improve gait initiation in people with Parkinson’s disease?,” Hum Mov Sci , vol. 84, Aug. 2022, doi: 10.1016/j.humov.2022.102970. [29] K. Poláková et al. , “3D visual cueing shortens the double support phase of the gait cycle in patients with advanced Parkinson’s disease treated with DBS of the STN,” PLoS One , vol. 15, no. 12 12, Dec. 2020, doi: 10.1371/journal.pone.0244676. [30] E. N. Beck, K. A. Ehgoetz Martens, and Q. J. Almeida, “Freezing of Gait in Parkinson’s Disease: An Overload Problem?,” PLoS One , vol. 10, no. 12, p. e0144986, Dec. 2015, doi: 10.1371/journal.pone.0144986. [31] L. Graham et al. , “Visual Exploration While Walking With and Without Visual Cues in Parkinson’s Disease: Freezer Versus Non-Freezer,” Neurorehabil Neural Repair , vol. 37, no. 10, pp. 734–743, Oct. 2023, doi: 10.1177/15459683231201149. [32] O. Gál et al. , “Pavement patterns can be designed to improve gait in Parkinson’s disease patients,” Movement Disorders , vol. 34, no. 12, pp. 1831–1838, Dec. 2019, doi: 10.1002/mds.27831. [33] C. Barthel et al. , “The laser shoes,” Neurology , vol. 90, no. 2, pp. e164–e171, Jan. 2018, doi: 10.1212/WNL.0000000000004795. 104 [34] C. G. D. W. A. L. S. C. D. G. Liang Tang, “Rhythmic laser cue is beneficial for improving gait performance and reducing freezing of turning in Parkinson’s disease patients with freezing of gait,” Int J Clin Exp Med , 2017, Accessed: Dec. 07, 2023. [Online]. Available: https://www.researchgate.net/publication/330290372_Original_Article_Rhythmic_laser_cue_i s_beneficial_for_improving_gait_performance_and_reducing_freezing_of_turning_in_Parkinson ’s_disease_patients_with_freezing_of_gait [35] J. D. Holmes et al. , “The Effects of Manipulating Spatial Location of Visual Cue Placement on Gait Among Individuals with Parkinson’s Disease: A Pilot Study,” Phys Occup Ther Geriatr , vol. 33, no. 3, pp. 263–278, Jul. 2015, doi: 10.3109/02703181.2015.1045109. [36] M. S. Bryant, D. H. Rintala, E. C. Lai, and E. J. Protas, “A pilot study: influence of visual cue color on freezing of gait in persons with Parkinson’s disease,” Disabil Rehabil Assist Technol , vol. 5, no. 6, pp. 456–461, Nov. 2010, doi: 10.3109/17483107.2010.495815. [37] L. Bunting-Perry, M. Spindler, K. M. Robinson, J. Noorigian, H. J. Cianci, and J. E. Duda, “Laser light visual cueing for freezing of gait in Parkinson disease:A pilot study with male participants,” The Journal of Rehabilitation Research and Development , vol. 50, no. 2, p. 223, 2013, doi: 10.1682/JRRD.2011.12.0255. [38] E. Cubo, C. G. Moore, S. Leurgans, and C. G. Goetz, “Wheeled and standard walkers in Parkinson’s disease patients with gait freezing,” Parkinsonism Relat Disord , vol. 10, no. 1, pp. 9– 14, Oct. 2003, doi: 10.1016/S1353-8020(03)00060-9. [39] C. Ratanasutiranont, K. Srisilpa, P. Termsarasab, and P. Ruthiraphong, “Smart ankle braceletlaser device to improve gait and detect freezing of gait in Parkinsonism patients: a case series,” Assistive Technology , vol. 35, no. 5, pp. 417–424, Sep. 2023, doi: 10.1080/10400435.2022.2113179. [40] E. J. An, W.-S. Sim, S. M. Kim, and J. Y. Kim, “Suitability of visual cues for freezing of gait in patients with idiopathic Parkinson’s disease: a case–control pilot study,” J Neuroeng Rehabil , vol. 20, no. 1, p. 91, Jul. 2023, doi: 10.1186/s12984-023-01214-8. [41] D. J. Geerse, B. Coolen, J. J. van Hilten, and M. Roerdink, “Holocue: A Wearable Holographic Cueing Application for Alleviating Freezing of Gait in Parkinson’s Disease,” Front Neurol , vol. 12, Jan. 2022, doi: 10.3389/fneur.2021.628388. [42] N. Byl, W. Zhang, S. Coo, and M. Tomizuka, “Clinical impact of gait training enhanced with visual kinematic biofeedback: Patients with Parkinson’s disease and patients stable post stroke.,” 111 ANOVA Step length [m] Square sum df Mean square Z Sig. Between groups .291 2 .145 6.193 .003 In the groups 3.100 132 .023 Total 3.391 134 Scheffe test Step length [m] (I) Study group (J) Study group Mean difference (I-J) Error Sig. Confidence level 95% Minimum limit Maximum limit 1 2 -.023061 .032143 .773 -.10264 .05652 3 -.107813* .031726 .004 -.18636 -.02927 2 1 .023061 .032143 .773 -.05652 .10264 3 -.084752* .033449 .044 -.16756 -.00194 3 1 .107813* .031726 .004 .02927 .18636 2 .084752* .033449 .044 .00194 .16756 *. A diferença média é significativa no nível 0.05. The distribution of Stance phase [%] is the same in the Study group categories .057 Retain null hypothesis The distribution of Swing phase [%] is the same in the Study group categories .044 Reject null hypothesis The distribution of Velocity [m/s] is the same in the Study group categories .008 Reject null hypothesis The distribution of Cadence [steps/min] is the same in the Study group categories .033 Reject null hypothesis The distribution of SD step length [m] is the same in the Study group categories .011 Reject null hypothesis The distribution of SD step time [s] is the same in the Study group categories .266 Retain null hypothesis The distribution of SD velocity [m/s] is the same in the Study group categories .008 Reject null hypothesis The distribution of AS step length [m] is the same in the Study group categories .006 Reject null hypothesis The distribution of AS step time [s] is the same in the Study group categories .027 Reject null hypothesis The distribution of AS velocity [m/s] is the same in the Study group categories .158 Retain null hypothesis 112 APPENDIX B – Custom-made cue perception Questionnaire Questions How perceptible was the laser? With what frequency did you use the laser cues? How perceptible was the vibration during the trial? How perceptible was the vibration in the first 20s of the trial? 1 - Not perceptible at all/did not use at all; 5 – Always perceptible/used it always