Enhancing epidemic management: agent-based simulation and remote diagnosis

Brazilian Journal of Technology 1 ISSN: 2595-5748 Brazilian Journal of Technology, Curitiba, v.7, n.2, p. 01-17, 2024 Enhancing epidemic management: agent-based simulation and remote diagnosis Melhorando o gerenciamento de epidemias: simulação baseada em agentes e diagnóstico remoto Mejora de la gestión de epidemias: simulación basada en agentes y diagnóstico remoto DOI:10.38152/bjtv7n2-006 Submitted: May 10st, 2024 Approved: May 31th, 2024 Djamel-Eddine Abdelaziz Graduated in Computer Science Institution: University of Batna 2 Address: Batna, Algeria E-mail: [email protected] Ouahab Kadri PhD in Industrial Engineering Institution: University of Batna 2 Address: Batna, Algeria E-mail: [email protected] ABSTRACT The objective of this work is to analyze the spread of contagious diseases using a multiagent simulation and a remote diagnostic system. The simulation is carried out using information on patients and the environment. The novelty of this work lies in proposing a new agent architecture that combines both independent and random behavior to better simulate individuals in the real world. We have proposed two communication methods between the agents either by a direct exchange of messages or by placing objects in a common space. The various simulations have enabled the prediction of the time required to transition from a normal state to a critical condition. A comparative study between the proposed diagnostic method and alternative techniques has demonstrated its superiority in accuracy. Keywords: contagious diseases, multi-agent model, artificial intelligence, reactive agent. RESUMO O objetivo deste trabalho é analisar a propagação de doenças contagiosas utilizando uma simulação multiagente e um sistema de diagnóstico remoto. A simulação é realizada a partir de informações do paciente e do ambiente. A novidade deste trabalho está em propor uma nova arquitetura de agentes que combina comportamento independente e aleatório para melhor simular indivíduos no mundo real. Propomos dois métodos de comunicação entre os agentes, seja por troca direta de mensagens ou pela colocação de objetos em um espaço comum. As diversas simulações permitiram prever o tempo necessário para a transição de um estado normal para uma condição crítica. Um estudo comparativo entre o método diagnóstico proposto e técnicas alternativas demonstrou sua superioridade em acurácia. Brazilian Journal of Technology 2 ISSN: 2595-5748 Brazilian Journal of Technology, Curitiba, v.7, n.2, p. 01-17, 2024 Palavras-chave: doenças contagiosas, modelo multiagente, inteligência artificial, agente reativo. RESUMEN El objetivo de este trabajo es analizar la propagación de enfermedades contagiosas mediante una simulación multiagente y un sistema de diagnóstico remoto. La simulación se lleva a cabo utilizando información sobre los pacientes y el medio ambiente. La novedad de este trabajo radica en proponer una nueva arquitectura de agentes que combina comportamiento independiente y aleatorio para simular mejor a los individuos en el mundo real. Hemos propuesto dos métodos de comunicación entre los agentes ya sea mediante un intercambio directo de mensajes o colocando objetos en un espacio común. Las distintas simulaciones han permitido predecir el tiempo necesario para pasar de un estado normal a una condición crítica. Un estudio comparativo entre el método de diagnóstico propuesto y técnicas alternativas ha demostrado su superioridad en precisión. Palabras clave: enfermedades contagiosas, modelo multiagente, inteligencia artificial, agente reactivo. 1 INTRODUCTION Today, malaria is considered among the most dangerous and contagious diseases. According to the latest statistics, there are more than 220 million patients suffering from malaria in Africa. This disease caused the death of 600,000 people in 2023. The main cause of death among children under the age of five is malaria. Malaria transmission is the bite of the female mosquito. A main cause is blood transfusion using unsterilized means. After contamination, the patient may have mild symptoms. If he does not consult a doctor, his state is degraded and he can undergo death in less than 24 hours. This disease represents a real challenge for the different countries of Africa. The lack of financial means makes it impossible to set up clinical centers in the isolated villages. The ideal solution is to create mobile cells that predict the propagation of the disease to find the areas that can be infected and to use innovative means for rapid identification of the disease (Gomes, 2022). IT simulation and artificial intelligence can play an important role in the fight against this epidemic (Andradige, 2024). These tools have proven their importance in several areas. The composition of these tools will create a platform that makes it possible to act effectively to surround the affected areas and then eliminate the disease. This work aims to provide two effective tools. The first is a decision support system that provides government information necessary to take correct actions. The second tool is a chatbot that can be used either by the doctor or the patient. This tool makes it possible to diagnose the patient based on the information provided (images and symptoms). On the computer side, several contributions have been implemented. One of the important Brazilian Journal of Technology 3 ISSN: 2595-5748 Brazilian Journal of Technology, Curitiba, v.7, n.2, p. 01-17, 2024 proposals is the design of a reactive multi-agent system that is characterized by its simplicity and innovative communication techniques. In the second part, we proposed a new deep learning architecture that combines the selection of parameters and transfer Learning. This solution uses a small database and offers a very high accuracy rate. The rest of this paper is structured as follows: the recent works concerning the management of infectious diseases are presented in section 2. The architecture of the proposed simulation model is described in section 3. The details of our diagnostic system are presented In section 4. Section 5 is devoted to showing the results of the various experiments followed by a discussion. In section 6, a general conclusion and some perspectives are presented. 2 RELATED WORK Simulation represents a very good alternative to data collection since the creation of a real database requires several months or years. In addition, the database created suffers from several problems such as missing or false values caused by damaged sensors. Thus, the simulation allows a predictive analysis with reduced costs thanks to an environment with several variables that are used to configure the different characteristics of the environment. In the majority of cases, simulation results are considered reliable and can be used to develop a predictive analysis to obtain recommendations to improve the studied system. At least, we can distinguish eight types of simulation models (compartment, agentbased, discrete-time, continuous-time, deterministic, stochastic, non-spatial, spatial gold). A careful study allows us to set the choice of the model to be used. In several cases, it is appropriate to use a hybrid model. A multi-agent system is made up of several entities. Each entity performs a set of actions that represents its behavior which is based on rules and interactions with the other entities. Depending on their behavior, we can distinguish two types of agents (reactive and cognitive). An agent is considered reactive if the actions follow the principle of causation. On the other hand, if the agent integrates an intelligence aspect such as an expert system or machine learning, we consider it cognitive. The multi-agent system can be considered to be intelligent even if all the agents are reactive. For example, the behavior of the colony is intelligent but each ant is based on the density of pheromone. In (Iwanaga, 2023), they implemented a model using a multi-agent system based on mobile cellular automata. To improve the interaction and behavior of agents, they proposed Brazilian Journal of Technology 4 ISSN: 2595-5748 Brazilian Journal of Technology, Curitiba, v.7, n.2, p. 01-17, 2024 methods to compare the parameters of the model with real geographic, social, and medical indicators. The developed model makes it possible to assess the impact of the quarantine, and restrictions on transport connections between regions, and to consider factors such as the incubation period, the mask regime, the maintenance of distance safety between people, etc. They developed a method of comparing time parameters and dynamics of the model with real data, which made it possible to assess the effectiveness of the government in stopping the pandemic in the region of Chernivtsi, in Ukraine. In (Niemann, 2024), they have developed a heterogeneous multi-level optimization approach combining an ABM at the end with a coarse level ode to find exemplary non-pharmaceutical interventions in the design of epidemic policy. Several works have been proposed concerning the diagnosis of infectious diseases based on images and conversations with the patient. In (Nahiduzzaman, 2023), they proposed a method of identifying computer-assisted pulmonary diseases based on CXR images. They used 17 different forms of pulmonary disorders which were divided into six sets of different forms of pulmonary disorders. The proposed frame combines the extraction of characteristics, the parallel CNN, and the ELM. In (Kaya, 2023), A model proposed to detect diseases on the images of the chest Xrays of individuals with pneumonia and healthy individuals. They have proposed a hybrid features extraction network, namely D3Senet, which consists of Darknet53, Darknet19, Densenet201, Squeezenet, and EfficientNetB0. Once a balanced data set has been prepared, the functional vectors have been obtained from images using CNN models based on deep learning and the size of the feature vectors has been reduced using a selection features method. The subset of features obtained had been classified by SVM. The number of features to be selected has been tested by the Iterative Incretion method and the features with the highest precision rate were obtained. Simulation and prediction tools focus on the final result and omit the progression of the system and the different intermediate stages. The transition from one step to another provides very useful information for the comprehension of the propagation of the disease. Another very important element is remote diagnosis which represents a significant source of data and an effective solution in countries with an insufficiency of medical staff. Brazilian Journal of Technology 5 ISSN: 2595-5748 Brazilian Journal of Technology, Curitiba, v.7, n.2, p. 01-17, 2024 3 MULTI-AGENT SIMULATION MODEL 3.1 NETLOGO The use of reactive agents allows the development of rapid simulation models. So, we have chosen the Netlogo platform which offers a set of tools that are easy to use. Netlogo was created in 1999 by Uri Wilensky at the Center for Connected Learning and ComputerModelization, then at Tofts University in Boston. The source of inspiration for Netlogo's developers is the Starlogot software that was created by Wilensky in 1997. In 2000, the continuation of improvements was made at Northwestern University, in the Chicago region. The initial objective of this tool is to provide developers with a complete programming environment for the creation of simulation models based on methods inspired by the real world. The code of any model relates to one or more agents. Agents can execute in parallel which is the case, for example, in groups of ants and bees. In Netlogo, we can identify four types of agents. The first type is called patch which is characterized by two coordinates in a two -dimensional space. The main agent is called Turtle which is created by the first type. Agent Turtle can move in the simulation space by performing several actions. The role of Agent Link is to ensure communication between different Turtles. The objective of the Observer agent is to supervise the simulation environment. This agent is unique and it can create, supervise, or destroy Turtles agents (BERGHOUT, 2020) (Rand, 2008) (Wilensky, 2015). 3.2 CLUSTERING MODEL The algorithm Lumer & Faieta is a variant of the algorithm proposed by Deneubourg. The first use of this algorithm was the classification of digital data. The various agents of this algorithm imitated the behavior of ants to sort the food. The ants use a function that calculates the similarity between the held object and the object on the ground to choose the action to be executed. The exploration of space by moving objects allows the colony to create classes of similar objects. By observing the result of the execution of this algorithm, we can conclude that the type of this classification is unsupervised. We used to calculate the similarity, of Minkowski's metric: 𝑑𝑟(𝑜𝑖;𝑜𝑗) = 𝑑𝑟(𝑥𝑖;𝑥𝑗)∑𝑀 𝑘=1 𝑊𝑘(|𝑥𝑖𝑘 − 𝑥𝑗𝑘|𝑟) 1 𝑟 (1) Brazilian Journal of Technology 6 ISSN: 2595-5748 Brazilian Journal of Technology, Curitiba, v.7, n.2, p. 01-17, 2024 We use the Hamming distance to calculate the distance between non-digital data. For example, to evaluate the distance between two character strings, we count the identical bits that have the same position (Anne, 2023). Figure 1. Clustering Model Source: The authors, 2024 3.3 PROPAGATION MODEL The proposed model illustrates the evolution of the infection between members of a society. The model interface comprises several scrolling bars, making it possible to modify the initial value of each parameter. The role of the main window is to show the results after the end of each iteration. The model offers the possibility of changing the configuration in the middle of the simulation. Each button allows one or more procedures to be triggered. The population creation procedure is carried out first. Then, it triggers the creation and configuration procedure for each agent. The choice of infected agents is made randomly but the rate of infected agents is defined by the user. We can increase or decrease the interaction between agents to illustrate the effects of social relations on the propagation or disappearance of the disease. We can distinguish two types of disease propagation. The first concerns the infection of individuals and the second concerns the regions which do not contain immune individuals. This second type represents a great threat to the health system of the infected region since generally, hospitals do not have logistical means to treat a huge number of patients at the same time. Before starting the simulation, we have to activate Configure button so that the system takes into account the initial configuration of the model. The stop criterion, which Brazilian Journal of Technology 7 ISSN: 2595-5748 Brazilian Journal of Technology, Curitiba, v.7, n.2, p. 01-17, 2024 completes the execution of the simulation, can be defined by a maximum number of iterations or the simulation ends after the infection of all agents. Another procedure is defined to update the illustration of the results. The last button is used to interrupt the execution of the model. We have created configuration interfaces for variables that have a direct influence on the propagation of infection. For example, we estimated that the degree of mobility of an agent has a great influence on the speed of spread. The run of several simulations will allow us to know this degree and in addition the influences of other parameters on the speed of propagation. The proposed model is expandable and it allows the addition of new parameters without modifying the majority of procedures. Netlogo offers a set of primitives that have facilitated the implementation of our model. The Best Method to configure a model or analyze the result is the use of CSV files. This format is widely used and supported by several data analysis software. Setting the initial values of parameters or defining a modification law makes it possible to limit their influence and focus on the relationship of results with each input of the model. Figure 2. Propagation Model Source: The authors, 2024 Brazilian Journal of Technology 8 ISSN: 2595-5748 Brazilian Journal of Technology, Curitiba, v.7, n.2, p. 01-17, 2024 4 DIAGNOSTIC SYSTEM 4.1 MEDICAL TEXT CLASSIFICATION USING CONVOLUTIONAL NEURAL NETWORKS The proposed Chatbot analyzes the text using linguistic and statistical methods to create models. The input text is transformed into manipulation structures. The text analysis process is made up of several stages. The most important are as follows: we decompose the text into small units, and then the irrelevant units are deleted. The most difficult step is to reduce words to their root. After converting units into digital values, we applied a CNN algorithm (Zhou, 2022). To implement our model, we used Flutter which is development kit software for open-source user interface created by Google. We also used Firebase MLKIT which is a database hosted by the cloud with data stored in the form of JSON. Part of our application was created using DART which is a programming language designed for customer development, as for web and mobile applications. It is developed by Google and can also be used to create server and desktop applications. And obviously, we have also used Django which is an open source web framework based on code reuse. Figure 3. Text Classification Architecture Source: The authors, 2024 Brazilian Journal of Technology 9 ISSN: 2595-5748 Brazilian Journal of Technology, Curitiba, v.7, n.2, p. 01-17, 2024 Figure 4. Use case diagram Source: The authors, 2024 4.2 IMAGE CLASSIFICATION 4.2.1 Architecture of the CNN The proposed diagnostic system uses convolutional neural networks to classify images. Our model is made up of several layers. We have chosen this model after a comparative study with the most used methods in the field of image classification. The essential characteristic of CNN is the use of tensors as a processing unit. Figure 5. Architecture of the CNN Source: The authors, 2024 Unlike standard neural networks, the parameter extraction step is included in CNN Brazilian Journal of Technology 16 ISSN: 2595-5748 Brazilian Journal of Technology, Curitiba, v.7, n.2, p. 01-17, 2024 REFERENCES ANDRADIGE, H. J., & LIYANAGE, L. (2024). Spatial distribution of noncommunicable diseases; a case of Ischaemic Heart Diseases, Cancer and Diabetes in Sri Lanka. Brazilian Journal of Development Vol. 10. 5, p. e69370. South Florida Publishing LLC. https://doi.org/10.34117/bjdv10n5-001 ANNE RAMYA, T., MATHANA, J. M., NIRMALA, R., & GOMATHI, R. (2023). Exploration on enhanced Quality of Services for MANET through modified Lumer and Fai-eta algorithm with modified AODV and DSR protocol. In Materials Today: Proceedings,Vol. 80, pp. 1765–1771. Elsevier BV. https://doi.org/10.1016/j.matpr.2021.05.601 BERGHOUT, T., MOUSS, L. H., KADRI, O., & HADJIDJ, N. (2020). Regularized Length Changeable Extreme Learning Machine with Incremental Learning Enhancements for Remaining Useful Life Prediction of Aircraft Engines. In 020 1ST INTERNATIONAL CONFERENCE ON COMMUNICATIONS, CONTROL SYSTEMS AND SIGNAL PROCESSING (CCSSP). 2020 1st International Conference on Communications, Control Systems and Signal Processing (CCSSP). IEEE. https://doi.org/10.1109/ccssp49278.2020.9151607 BHANDARY, A., PRABHU, G. A., RAJINIKANTH, V., THANARAJ, K. P., SATAPATHY, S. C., ROBBINS, D. E., SHASKY, C., ZHANG, Y.-D., TAVARES, J. M. R. S., & RAJA, N. S. M. (2020). Deep-learning framework to detect lung abnormality – A study with chest X-Ray and lung CT scan images. Pattern Recognition Letters,Vol. 129, pp. 271–278. Elsevier BV. https://doi.org/10.1016/j.patrec.2019.11.013 DHRUV, P., & NASKAR, S. (2020). Image Classification Using Convolutional Neural Network (CNN) and Recurrent Neural Network (RNN): A Review. Machine Learning and Information Processing pp. 367–381. Springer Singapore. https://doi.org/10.1007/978-981-15-1884-3_34 EISELE, T. P., LARSEN, D. A., ANGLEWICZ, P. A., KEATING, J., YUKICH, J., BENNETT, A., HUTCHINSON, P., & STEKETEE, R. W. (2012). Malaria prevention in pregnancy, birthweight, and neonatal mortality: a meta-analysis of 32 national crosssectional datasets in Africa. The Lancet Infectious Diseases,Vol. 12, Issue 12, pp. 942– 949. Elsevier BV. https://doi.org/10.1016/s1473-3099(12)70222-0 GOMES, A. P., BRAGA, L. M., FERREIRA, J. A., VICARI, M. V., GOMES, V., & MOTTA, O. J. R. DA. (2022). Biosecurity and infectious diseases: contemporary challenges / Biossegurança e doenças infecciosas: desafios contemporâneos. Brazilian Journal of Health Review,Vol. 5. 2, pp. 6364–6391). South Florida Publishing LLC. https://doi.org/10.34119/bjhrv5n2-211 IWANAGA, S. (2023) Multiagent Simulation of Seasonal Influenza from Preinfectious People in Closed Spaces. Vietnam Journal of Computer Science ,Vol. 11, Issue 01, pp. 95–109, 2023. World Scientific Pub Co Pte Ltd. https://doi.org/10.1142/s219688882340002x KAYA, M., & ERIS, M. (2023). D3SENet: A hybrid deep feature extraction network for Brazilian Journal of Technology 17 ISSN: 2595-5748 Brazilian Journal of Technology, Curitiba, v.7, n.2, p. 01-17, 2024 Covid-19 classification using chest X-ray images. Biomedical Signal Processing and Control ,Vol. 82, p. 104559. Elsevier BV. https://doi.org/10.1016/j.bspc.2022.104559 NIEMANN, J.-H., URAM, S., WOLF, S., DJURDJEVAC CONRAD, N., & WEISER, M. (2024). Multilevel optimization for policy design with agent-based epidemic models. Journal of Computational Science, Vol. 77, p. 102242. Elsevier BV. https://doi.org/10.1016/j.jocs.2024.102242 NAHIDUZZAMAN, Md., GONI, Md. O. F., HASSAN, R., ISLAM, Md. R., SYFULLAH, M. K., SHAHRIAR, S. M., ANOWER, Md. S., AHSAN, M., HAIDER, J., & KOWALSKI, M. (2023). Parallel CNN-ELM: A multiclass classification of chest X-ray images to identify seventeen lung diseases including COVID-19. Expert Systems with Applications ,Vol. 229, p. 120528. Elsevier BV. https://doi.org/10.1016/j.eswa.2023.120528 RAND, W., WILENSKY, U. (2008). NetLogo Spread of Disease model. http://ccl.northwestern.edu/netlogo/models/SpreadofDisease. Center for Connected Learning and Computer-Based Modeling, Northwestern Institute on Complex Systems, Northwestern University, Evanston, IL. WILENSKY, U. & RAND, W. (2015). Introduction to Agent-Based Modeling: Modeling Natural, Social and Engineered Complex Systems with NetLogo. Cambridge, MA. MIT Press. ZHOU, Y., Li, J., CHI, J., TANG, W., & ZHENG, Y. (2022). Set-CNN: A text convolutional neural network based on semantic extension for short text classification. Knowledge-Based Systems,Vol. 257, p. 109948. Elsevier BV. https://doi.org/10.1016/j.knosys.2022.109948