Artificial intelligence in medicine and healthcare: a review and classification of current and near-future applications and their ethical and social Impact

arXiv:2001.09778,2020. 1 Artificial intelligence in medicine and healthcare: a review and classification of current and near-future applications and their ethical and social Impact. Emilio Gómez-González1,2,3, Emilia Gomez4,5, Javier Márquez-Rivas6,7,2,1, Manuel Guerrero-Claro1, Isabel Fernández-Lizaranzu8,1, María Isabel Relimpio-López9,10, Manuel E. Dorado11, María José Mayorga- Buiza12,2,1, Guillermo Izquierdo-Ayuso13,14, Luis Capitán-Morales15. 1Group of Interdisciplinary Physics, Department of Applied Physics III, ETSI Engineering School, Universidad de Sevilla. Camino de los Descubrimientos sn, 41092 Seville, Spain. 2Group of Applied Neuroscience, Institute of Biomedicine of Seville (IBIS). Avda. Manuel Siurot sn, 41013 Seville, Spain. 3Royal Academy of Medicine and Surgery of Seville. C/. Abades 10, 41004 Seville, Spain. 4Joint Research Center, European Commission*. C/. Inca Garcilaso 3, 41092 Seville, Spain. 5Universitat Pompeu Fabra. Plaza de la Mercé 10, 08002 Barcelona, Spain. 6Service of Neurosurgery, Hospital Universitario “Virgen del Rocio”. Avda. Manuel Siurot sn, 41013 Seville, Spain. 7Center for Advanced Neurology (CNA). Avda. Manuel Siurot 43A, 41013 Seville, Spain. 8SeLIVER Group, Institute of Biomedicine of Seville (IBIS). Avda. Manuel Siurot sn, 41013 Seville, Spain. 9Department of Ophthalmology, Hospital Universitario “Virgen Macarena”. C/. Dr. Fedriani 3, 41009 Seville, Spain. 10RETICS, OftaRed, Instituto de Salud Carlos III (ISCIII). Avda. Monforte de Lemos 5, 28029 Madrid, Spain. 11Department of Human Anatomy and Embryology, College of Medicine, Universidad de Sevilla. Avda. Sánchez Pizjuán sn, 41009 Seville, Spain. 12Service of Anesthesiology, Hospital Universitario “Virgen del Rocio”. Avda. Manuel Siurot sn, 41013 Seville, Spain. 13Service of Neurology, Hospital “VITHAS-NISA Sevilla”. Avda. Plácido Fernández Viagas sn, 41950 Castilleja de la Cuesta, Seville, Spain. 14Service of Neurology, DINAC Foundation. C/. Fernando Álvarez de Toledo 4 2º-3, 41009 Seville, Spain. 15Department of Surgery, College of Medicine, Universidad de Sevilla. Avda. Sánchez Pizjuán sn, 41009 Seville, Spain. *Disclaimer: The content of this article does not represent the official position of the European Commission. Corresponding author: Prof.Dr. Emilio Gómez-González {egom[email protected]} Group of Interdisciplinary Physics, Department of Applied Physics III ETSI Engineering School - Universidad de Sevilla Camino de los Descubrimientos sn 41092 Sevilla, Spain www.etsi.us.es/gfi To cite this document: Gomez-Gonzalez E, Gomez E, Marquez-Rivas et al. Artificial Intelligence in Medicine and Healthcare: a review and classification of current and near-future applications and their ethical and social impact. arXiv:2001.09778,2020. There is a Spanish translation available for download at: https://5a8fe19c-a1f8-41fe-938d-8c3f23f975c9.filesusr.com/ugd/d42f6e_2fc902643d594faa8e80d82d7f43d2b6.pdf arXiv:2001.09778,2020. 2 Abstract This paper provides an overview of the current and near-future applications of Artificial Intelligence (AI) in Medicine and Health Care and presents a classification according to their ethical and societal aspects, potential benefits and pitfalls, and issues that can be considered controversial and are not deeply discussed in the literature. This work is based on an analysis of the state of the art of research and technology, including existing software, personal monitoring devices, genetic tests and editing tools, personalized digital models, online platforms, augmented reality devices, and surgical and companion robotics. Motivated by our review, we present and describe the notion of “extended personalized medicine”, we then review existing applications of AI in medicine and healthcare and explore the public perception of medical AI systems, and how they show, simultaneously, extraordinary opportunities and drawbacks that even question fundamental medical concepts. Many of these topics coincide with urgent priorities recently defined by the World Health Organization for the coming decade. In addition, we study the transformations of the roles of doctors and patients in an age of ubiquitous information, identify the risk of a division of Medicine into “fake-based”, “patient-generated”, and “scientifically tailored”, and draw the attention of some aspects that need further thorough analysis and public debate. Introduction Artificial Intelligence (AI) is a new realm of science and technology. It already affects many human activities at all societal levels, from individuals to social groups, corporations and nations. AI is expanding rapidly, worldwide, in almost every industrial, economical and societal sector, from information technologies to commerce, manufacturing, space, remote sensing, security and defense, transport and vehicles and, since the beginning of the XXI century, it is effectively entering into Medicine and Health Care1,2. The fast and powerful evolution of AI is fostered by many factors. Among them,  the availability of powerful and cost-effective computing (processing) tools, hardware (e.g. graphics processing units), software and applications, –even in consumer-grade personal computers and mobile devices– and of large (big) data sets, with many different types and formats of information, both in online and cloud platforms and generated in real time by user wearables and the internet-of-things (IoT),  the expansion of open source coding resources and online communities of users and practitioners sharing resources, expertise (know-how), and experience, and  the combination of computer processing with other technologies such as photonics (merging of applied optics and electronics) and human-machine interfaces. Recent advances in AI systems in Medicine and Health Care present extraordinary opportunities, – particularly in areas of such deep social interest as oncology3– together with significant questions4 and drawbacks5–7, calling for a close consideration of their implementation8 and how they affect – and can even change– basic definitions in the medical context9–11. This study provides a review of existing and near-future applications of AI in this particular sector from the point of view of their potential benefits and pitfalls, ethical and social impact. We also arXiv:2001.09778,2020. 3 identify a set of controversial issues that are not deeply discussed in the literature and should be further researched. Notably, many of the topics presented in this work coincide with six of the thirteen urgent priorities recently defined by the World Health Organization (WHO) for the coming decade12. These coinciding priorities explicitly include: “Harnessing new technologies”, “Earning public trust”, “Protecting people from dangerous products”, “Making health care fairer”, “Expanding access to medicines”, and “Preparing for epidemics”. Within the specific priority of “Harnessing new technologies”, the WHO defines the challenge as “New technologies are revolutionizing our ability to prevent, diagnose and treat many diseases. Genome editing, synthetic biology and digital health technologies such as artificial intelligence can solve many problems, but also raise new questions and challenges for monitoring and regulation. Without a deeper understanding of their ethical and social implications, these new technologies, which include the capacity to create new organisms, could harm the people they are intended to help”. This work is based on an exhaustive literature review of existing scientific publications and technical publications, including software packages, personal monitoring devices, genetic tests and editing tools, personalized digital models, online platforms, augmented reality devices, and surgical and companion robotics13. This review covered 588 references, most of them from 2016-2019 but intended to be dynamically updated, including standard scientific and academic platforms such as MEDLINE, Current Contents and PubMed plus product descriptions and internet and press articles of well-recognized sources which are relevant for the topic. From this review, we summarize here our main findings, discussion points and conclusions, supported by 107 representative references. “Tech-savvy”: From early pioneers in imaging and surgery to robotics and online platforms. In Medicine and Health Care, AI has evolved from computer programs to support the analysis of medical images to its integration in almost every clinical14 and organizational area15. Radiology16 was at the forefront of this transformation, together with different branches of surgery using augmented reality devices17 and surgical robots18. They were quickly followed by other image-based specialties (e.g. pathology19, dermatology20, ophtalmology21) and, more recently, by virtually all areas of Medicine and Health Care, from general practitioners22 to emergency departments23, epidemiology24, and disease management25. Systems for “computer-aided diagnosis” have expanded to include online assistants (e.g. app, chatbots) –both for very specific medical areas (e.g. oncology26, predicting the response to treatments27) and for the general public28–, and “clinical robots” now include “social companions” for hospitalized person, particularly children29 and the elderly30. In addition, wearables and IoT devices allow for real-time monitoring of physiological information, even at home31, and, integrated with medical and social-media data, can trigger clinically related automated interventions (from suicide prevention calls to police32 to medication delivery33). Concerned –even healthy– citizens can now order direct-to-consumer genetic tests among many thousands in the market34. New tools for big data modeling, analysis, and visualization are also expanding, and provide substantial, transforming improvements in clinical pathways35, from the generation of “digital twins”36 of individual patients to self-management of treatments37. There are even online, crowd shared platforms for such high-end applications as radiotherapy38. Many management aspects related to health economy (e.g. increased efficiency, quality control, fraud reduction) and policy39 also benefit from the new AI mediated tools. They even offer new hopes of improvements in health for environments with reduced resources40 and in developing regions41. arXiv:2001.09778,2020. 4 However, technical challenges42–44 and ethical concerns45–47 remain, and new important questions arise. Classification of the applications of Artificial Intelligence in Medicine and Health Care. In Figure 1 we propose a graded classification of AI and AI-mediated applications in Medicine and Health Care according to their beneficial vs detrimental character as recognized in the reviewed literature. For each of these applications, we report the technology, the specific implementations behind and their level of availability according to published references. On the one side of the spectrum of social impact we consider applications that have been shown to be beneficial to support clinical decision making, while on the other side we place applications that are widely agreed to be harmful. In Figure 1 we also propose a (new) scale named "Technology Availability Level" (TAL) which gives a qualitative description of the degree of availability of a technology, in a numerical scale from 0 (unknown) to 9 (available for the general public). The TAL scale is similar in format (and related) to the standard "Technology Readiness Levels" (TRL) but, as mentioned, it is based on published references (in scientific and academic literature, industrial or corporate reports, and general media citing sources considered to be reliable according to standards). It is important to consider that "availability" is not necessarily equivalent to "readiness levels" due to such factors as disclosure according to industrial, proprietary and/or government strategies, and that the TAL scale does not evaluate either the fulfillment of regulatory processes. The values defined for the TAL scale are the following: TAL 0. Unknown status. Not considered feasible according to references. TAL 1. Unknown status. Considered feasible according to related, indirect references. TAL 2. General/basic idea publicly proposed. TAL 3. Calls for public funding of R&D open. TAL 4. Results of academic/partial projects disclosed. TAL 5. Early design of product disclosed. TAL 6. Operational prototype/'first case' disclosed. TAL 7. Products disclosed but not available. TAL 8. Available for restricted (e.g. professional) users. TAL 9. Available for the public. arXiv:2001.09778,2020. 5 Figure 1. Proposal of classification of artificial intelligence (AI) and AI-mediated applications in Medicine and Health Care according to their beneficial vs detrimental character. SW = software, AR = augmented reality, VR = virtual reality, IoT = internet of things. TAL = Technology Availability Level (see details of values in the text). arXiv:2001.09778,2020. 6 In Figure 2 we complement this classification with an analysis of the main related (ethical and social) aspects of AI technologies by defining three partially overlapping sets: 1. In the first group we include those topics commonly under analysis as raised by other AI application domains (e.g. social networks, online commerce, automation in factories, autonomous vehicles) such as data use and sharing48,49, privacy50, security, and annonymization51. 2. In the second group, we consider topics –some of which are also under analysis in other sectors– which are of particular relevance in our context. Among them, the proper training of AI systems to avoid bias52 and assure fairness and equity53, the difficulties to explain the “reasoning process” of complex systems54, the lack of updated evaluation protocols55,56 and regulatory standards57,58, the legal aspects of liability and malfunction59, and the vulnerability of medical information under adversarial attacks60. And, of course, the question of whether (or not) harnessing AI systems under human control61. 3. Finally, in the third group we place some aspects barely or not included in the reviewed literature. They mainly relate to the potential dual use of certain technologies, from limitdefying research about human control, ‘biological robots’, human-animals hybrids and biohacking to weaponization and bioterrorism. They will be discussed –with examples from neuroscience and gene editing– in a later section. It is important to note that many AI systems and AI-mediated applications show an intrinsic “mix” of positive, negative, and controversial aspects depending on their specific implementations, and that, according to published information, their readiness levels vary from commercially available to very early, conceptual designs. To our knowledge, this is the first attempt to provide an exhaustive list of AI systems and applications in Medicine and Health Care taking into account their potential benefits and pitfalls. However, it is not intended to define an “absolute” scale of “goodness” or “badness”, as many technologies (e.g. gene editing, neuroprostheses) are not necessarily “positive” or “negative”, and others may certainly be difficult to categorize. Even the scientific and ethical criteria for such analysis need a review, and new questions arise. The methods used to evaluate the performance of medical products and treatments are based on averages over population groups, while we now consider extremely personalized approaches, to the genetic structure of each individual. How can those treatments be rigorously tested? Which are the time and cost required to find “enough cases” to “generate scientific evidence”? In addition, how should AI systems be benchmarked? Should they be compared to a (possibly error-prone) human doctor or “against” another “machine”? Common ethical guidelines for the evaluation of technologies mostly date from the pre-digital era. Which should be the figures of merit to consider? Which are the roles of the public and the policy makers? The goal of the classification proposed in these figures is to draw attention to the need of a careful, thorough analysis and as the basis for our discussion below, with a focus on those issues hardly addressed in the current literature. As mentioned, it is particularly important that many of these topics have been explicitly defined as “urgent priorities for the coming decade” by the WHO12 at the beginning of 2020. arXiv:2001.09778,2020. 7 Figure 2. Ethical and social aspects of artificial intelligence (AI) and AI-mediated technologies in Medicine and Health Care in three groups (G1, G2, and G3). Some key relevant questions, controversies, significant, and conflicting issues are outlined for each aspect. arXiv:2001.09778,2020. 8 “Extended Personalized Medicine”. The original goal of Personalized Medicine is to exploit very specific biological (genetic) features of individuals for tailored diagnosis and treatment62. Decoding the genome of each patient represents a very significant change from the existing model of averaged analysis of populations to an extremely individualized approach63, for treating disease –in a new paradigm defined as “precision medicine”64– but also to promote wellness65 and healthy, personalized lifestyles. However, although not explicitly formulated in the literature, we consider that the underlying principle of Personalized Medicine can be further expanded. It can include other properties whose particular values or structures –even their spatial distribution and time evolution in the human body– may be significantly different for any single individual, in different clinical situations, at every moment of life and, possibly in strong relationship to each other. Such additional features form the new concept of “Extended Personalized Medicine” and they may come from  “known sources” from the “basic sciences” of physics (e.g. bioelectromagnetic fields and signals, biomechanical magnitudes and properties, hydrodynamic parameters of the circulation of any fluid in the body), chemistry (concentrations of ions, molecules), and biology (metabolites),  “not yet known” origins. This concept refers to the potential characterization of brain processing schemes, connections, and functions whose details still remain veiled for science,  demographic data, extracted from conventional databases,  data about psychological and emotional status, extracted indirectly from the individual activity in social networks.  social data, including those about societal structures (family, groups providing psychological, emotional support), and cultural and religious beliefs which may influence health-related issues, such as restrictions on types of food or sexual activity, provided by the user or mined from social networks,  “lifestyle parameters” (sleep hours, stress, physical activity, food ingest) easily accessible through apps, wearables, and IoT,  values of environmental and physical geography conditions (weather, contamination) transmitted by multiple platforms,  sensors evaluating mood through face and gesture recognition, changes in cardiac rhythms, perspiration, and breathing patterns when receiving certain visual or auditory stimuli. They may be biometrics readers in smartphones, domotic environments, and wearables, Many sources of such data are already available. They include “standard” clinical scores and evaluations (imaging scans, analytics), results of genetic and “omics” (genomics, metabolomics, proteomics) tests, and medical knowledge from publications and references, but also data from social networks66 and from personal, IoT, and home sensors67. Together with potential benefits68, such as predicting individual response to treatments69, this “accumulation” of personal, intimate information also presents very high risks regarding ownership, security and privacy. From a technical point of view, the combination of varied heterogeneous sources requires the use of advanced AI-mediated tools for merging, processing, and analyzing multiple data layers – extracting useful information– and to operate the devices of augmented and virtual reality for the (very much needed) interactive navigation and visualization. arXiv:2001.09778,2020. 9 Replacement or enhancement: Do we need a doctor? AI systems pose important conceptual challenges for physicians. The most relevant is the debate between AI-replaced or AI-enhanced doctors70,71. As in any other sector, the appearance of (even partially) automated systems produces deep professional transformations, as some jobs will be lost and new ones will arise. Current systems for clinical applications may be mostly considered as “weak AI” that is, they can succeed at “narrow” tasks and be useful helping human physicians in tedious, repetitive activities (e.g. finding elements or segmenting organs or structures in images). As mentioned, medical specialties dealing with visual analysis of images have started to experiment substantial changes through such “aids”. Bottom-line is that computerized systems can process many, many more images than any human operator can, without fatigue and time constraints, and –if properly trained– with even better results. AI systems also show very good performance in the recognition of natural language and written texts, which in turn opens the way to extracting information –for systems to learn– from many available sources (e.g. clinical records) which are not usually available in structured, useful formats. But in general, medical AI systems lack the ability to “interpret” the context and “generate” the most distinctive human features (creativity, emotions), and cannot be considered (yet) as “strong” or ”general” AI72. Although many questions remain73, AI systems can enhance many aspects of doctor’s tasks and those professionals choosing not to use them will be outdated. The availability of massive amounts of data conveying very different types of information of each individual patient also requires new professional profiles –not necessarily physicians– to explore and extract the most useful items for diagnosis and treatment. Genetic counselors are already joining clinical teams to help (patients but also professionals) in understanding the complex generelated information74. The addition of novel sources of data –in the foreseeable paradigm of “expanded personalized medicine”– will likely require some other profiles of “medical data scientist” for specialized advice for physicians and patients. Key intangible qualities. Healthcare is not just an intellectual interaction but highly social, especially relevant in situations of vital relevance (e.g. decisions with effects on life quality, palliative care, euthanasia). There are many different views on trust, empathy, and the humanization of care, and how they can be affected by the new reality of automated systems, from the reduction of medical errors to the possibility of AI- systems helping humans to take difficult decisions, to the design of better doctor-AI and patient-AI interaction strategies75,76. Proponents of AI-enhanced human doctors see the integration of new tools as a powerful way to regain human aspects of the relationship between the doctor and the patient77 but the real results are yet unknown78. A key issue in of the greatest importance in all AI-related areas is that of increasingly autonomous systems79. In the medical and surgical fields, their capabilities may include the decision about human lives and, if they ever become available, which would be their ethical guidelines? For many people, this scenario may seem unreal, but there is a current debate about a topic that shares some of the arguments and positions, namely that about lethal autonomous weapons systems. Such devices have already made their way to the public –even to be discussed at the United Nations80 – as this type of research has been (at least) partially disclosed and early systems begin to be displayed. Their objectives are clearly the opposite of medical devices, and the popular name of “killer robots” may ward it off from being included in medical literature, but the fundamental idea to discuss is the same: will “a machine” take the ultimate decision to keep or end a human life? arXiv:2001.09778,2020. 16 Oncology 2019; 20: e253--e261. 20 Schlessinger DI, Chhor G, Gevaert O, Swetter SM, Ko J, Novoa RA. Artificial intelligence and dermatology: opportunities, challenges, and future directions. Seminars in Cutaneous Medicine and Surgery 2019; 38: E31–7. 21 Ting DSW, Lee AY, Wong TY. An Ophthalmologist’s Guide to Deciphering Studies in Artificial Intelligence. Ophthalmology 2019; 126: 1475–9. 22 Blease C, Bernstein MH, Gaab J, et al. Computerization and the future of primary care: A survey of general practitioners in the UK. PLOS ONE 2018; 13: e0207418. 23 Berlyand Y, Raja AS, Dorner SC, et al. How artificial intelligence could transform emergency department operations. American Journal of Emergency Medicine 2018; 36: 1515–7. 24 Wiens J, Shenoy ES. Machine Learning for Healthcare: On the Verge of a Major Shift in Healthcare Epidemiology. Clinical Infectious Diseases 2018; 66: 149–53. 25 Feng S, Grépin KA, Chunara R. Tracking health seeking behavior during an Ebola outbreak via mobile phones and SMS. npj Digital Medicine 2018; 1: 51. 26 Jacobs C, van Ginneken B. Google’s lung cancer AI: a promising tool that needs further validation. Nature Reviews Clinical Oncology 2019; 16: 532–3. 27 Xu Y, Hosny A, Zeleznik R, et al. Deep Learning Predicts Lung Cancer Treatment Response from Serial Medical Imaging. Clinical Cancer Research 2019; 25: 3266–75. 28 Greene A, Greene CC, Greene C. Artificial intelligence, chatbots, and the future of medicine. The Lancet Oncology 2019; 20: 481–2. 29 Logan DE, Breazeal C, Goodwin MS, et al. Social Robots for Hospitalized Children. Pediatrics 2019; 144: e20181511. 30 Shishehgar M, Kerr D, Blake J. A systematic review of research into how robotic technology can help older people. Smart Health 2018; 7–8: 1–18. 31 Blott J. Smart homes for the future of dementia care. The Lancet Neurology 2019; 4422: 30249. 32 Coppersmith G, Leary R, Crutchley P, Fine A. Natural Language Processing of Social Media as Screening for Suicide Risk. Biomedical informatics insights 2018; 10: 1178222618792860. 33 Westman N. Apple Watches may soon decide when to administer medications. Popular Science. 2019. https://www.popsci.com/apple-watch-activity-tracker-dementia-treatment/ (accessed Oct 1, 2019). 34 Regalado A. 2017 was the year consumer DNA testing blew up. MIT Technology Review. 2018. https://www.technologyreview.com/s/610233/2017-was-the-year-consumer-dna-testing-blew-up/ (accessed Oct 15, 2019). 35 Azria D, Rosenstein BS. Use of genomics to balance cure and complications. Nature Reviews Clinical Oncology 2019; 17: 9–10. 36 Bruynseels K, Santoni de Sio F, van den Hoven J. Digital Twins in Health Care: Ethical Implications of an Emerging Engineering Paradigm. Frontiers in Genetics 2018; 9: 1–11. arXiv:2001.09778,2020. 17 37 Landmesser U, MacRae CA. Digital technology to support self-management in patients with coronary disease. The Lancet Digital Health 2019; 1: e50--e51. 38 Mak RH, Endres MG, Paik JH, et al. Use of Crowd Innovation to Develop an Artificial Intelligence– Based Solution for Radiation Therapy Targeting. JAMA Oncology 2019; 5: 654. 39 Ashrafian H, Darzi A. Transforming health policy through machine learning. PLOS Medicine 2018; 15: 1–3. 40 Wahl B, Cossy-Gantner A, Germann S, Schwalbe NR. Artificial intelligence (AI) and global health: how can AI contribute to health in resource-poor settings? BMJ Global Health 2018; 3: e000798. 41 Artificial intelligence in global health: a brave new world. The Lancet 2019; 393: 1478. 42 Saria S, Butte A, Sheikh A. Better medicine through machine learning: What’s real, and what’s artificial? PLOS Medicine 2018; 15: e1002721. 43 Hutson M. Artificial intelligence faces reproducibility crisis. Science 2018; 359: 725–6. 44 Maddox TM, Rumsfeld JS, Payne PRO. Questions for Artificial Intelligence in Health Care. JAMA - Journal of the American Medical Association 2019; 321: 31–2. 45 Next generation public health: towards precision and fairness. The Lancet Public Health 2019; 4: e209. 46 Vayena E, Blasimme A, Cohen IG. Machine learning in medicine: Addressing ethical challenges. PLOS Medicine 2018; 15: e1002689. 47 Weintraub A. Artificial Intelligence Is Infiltrating Medicine - But Is It Ethical? Forbes. 2018. https://www.forbes.com/sites/arleneweintraub/2018/03/16/artificial-intelligence-is-infiltrating- medicine-but-is-it-ethical/#715ad7783a24 (accessed July 25, 2019). 48 Grundy Q, Chiu K, Held F, Continella A, Bero L, Holz R. Data sharing practices of medicines related apps and the mobile ecosystem: traffic, content, and network analysis. BMJ 2019; 364: l920. 49 Ghassemi M, Naumann T, Schulam P, Beam AL, Chen IY, Ranganath R. Practical guidance on artificial intelligence for health-care data. The Lancet Digital Health 2019; 1: e157–9. 50 Price WN, Cohen IG. Privacy in the age of medical big data. Nature Medicine 2019; 25: 37–43. 51 Na L, Yang C, Lo C-C, Zhao F, Fukuoka Y, Aswani A. Feasibility of Reidentifying Individuals in Large National Physical Activity Data Sets From Which Protected Health Information Has Been Removed With Use of Machine Learning. JAMA Network Open 2018; 1: e186040. 52 There is no such thing as race in health-care algorithms. The Lancet Digital Health 2019; 1: e375. 53 Rajkomar A, Hardt M, Howell MD, Corrado G, Chin MH. Ensuring Fairness in Machine Learning to Advance Health Equity. Annals of Internal Medicine 2018; 169: 866. 54 Watson DS, Krutzinna J, Bruce IN, et al. Clinical applications of machine learning algorithms: beyond the black box. BMJ 2019; 364: l886. 55 Smallman M. Policies designed for drugs won’t work for AI. Nature 2019; 567: 7. 56 Anderson M, Anderson SL. How Should AI Be Developed, Validated, and Implemented in Patient arXiv:2001.09778,2020. 18 Care? AMA Journal of Ethics 2019; 21: E125-130. 57 Parikh RB, Obermeyer Z, Navathe AS. Regulation of predictive analytics in medicine. Science 2019; 363: 810–2. 58 Pesapane F, Volonté C, Codari M, Sardanelli F. Artificial intelligence as a medical device in radiology: ethical and regulatory issues in Europe and the United States. Insights into Imaging 2018; 9: 745–53. 59 Prabhu SP. Ethical challenges of machine learning and deep learning algorithms. The Lancet Oncology 2019; 20: 621–2. 60 Finlayson SG, Bowers JD, Ito J, Zittrain JL, Beam AL, Kohane IS. Adversarial attacks on medical machine learning. Science 2019; 363: 1287–9. 61 Taddeo M, Floridi L. How AI can be a force for good. Science 2018; 361: 751–2. 62 Abrahams E. Personalized Medicine: Creating an Ecosystem of Shared Value. Personalized Medicine in Brief 2019; 12. 63 Gambhir SS, Ge TJ, Vermesh O, Spitler R. Toward achieving precision health. Science Translational Medicine 2018; 10: eaao3612. 64 Lasalvia L, Merges R. Expanding Precision Medicine. Journal of Precision Medicine 2019; 5: 1–5. 65 Juengst ET, McGowan ML. Why Does the Shift from “Personalized Medicine” to “Precision Health” and “Wellness Genomics” Matter? AMA Journal of Ethics 2018; 20: E881-890. 66 Fiumara G, Celesti A, Galletta A, Carnevale L, Villari M. Applying Artificial Intelligence in Healthcare Social Networks to Identity Critical Issues in Patients’ Posts. In: Proceedings of the 11th International Joint Conference on Biomedical Engineering Systems and Technologies. SCITEPRESS - Science and Technology Publications, 2018: 680–7. 67 Rogers J, Malliaras G, Someya T. Biomedical devices go wild. Science Advances 2018; 4: 2–4. 68 Snyder M, Zhou W. Big data and health. The Lancet Digital Health 2019; 1: e252–4. 69 Turki T, Wang JTL. Clinical intelligence: New machine learning techniques for predicting clinical drug response. Computers in Biology and Medicine 2019; 107: 302–22. 70 Norgeot B, Glicksberg BS, Butte AJ. A call for deep-learning healthcare. Nature Medicine 2019; 25: 14–8. 71 Digital health: the good, the bad, and the abandoned. The Lancet Psychiatry 2019; 6: 273. 72 Boden M. Artificial Intelligence. A very short introduction. Oxford: Oxford University Press, 2018. 73 Matuchansky C. Deep medicine, artificial intelligence, and the practising clinician. The Lancet 2019; 394: 736. 74 Kilbride MK, Domchek SM, Bradbury AR. Ethical Implications of Direct-to-Consumer Hereditary Cancer Tests. JAMA Oncology 2018; 4: 1327. 75 Ofri D. Empathy in the age of the electronic medical record. The Lancet 2019; 394: 822–3. 76 Heaven D. Your next doctor’s appointment might be with an AI. MIT Technology Review. 2018. arXiv:2001.09778,2020. 19 https://www.technologyreview.com/s/612267/your-next-doctors-appointment-might-be-with-an- ai/ (accessed Aug 13, 2019). 77 Insel TR. How algorithms could bring empathy back to medicine. Nature 2019; 567: 172–3. 78 Topol EJ. High-performance medicine: the convergence of human and artificial intelligence. Nature Medicine 2019; 25: 44–56. 79 Statement on Artificial Intelligence, Robotics and ‘Autonomous’ Systems. European Commission. 2018. http://ec.europa.eu/research/ege/pdf/ege_ai_statement_2018.pdf (accessed July 25, 2019). 80 2019 Group of Governmental Experts on Lethal Autonomous Weapons Systems (LAWS). United Nations. 2019. https://www.unog.ch/80256EE600585943/(httpPages)/5535B644C2AE8F28C1258433002BBF14?Op enDocument (accessed Oct 1, 2019). 81 Topol E. The patient will see you now: The Future of Medicine is in your hands. New York: Basic Books, 2016. 82 Bose R, Saxon LA. The Democratization of Diagnosis: Bringing the Power of Medical Diagnosis to the Masses. EClinicalMedicine 2019; 8: 6–7. 83 Nichols T. The Death of Expertise: The Campaign Against Established Knowledge and Why it Matters. Oxford University Press, 2017. 84 Dzau VJ, Balatbat CA. Health and societal implications of medical and technological advances. Science Translational Medicine 2018; 10: 1–4. 85 Wells CJ. Emerging Gene Therapies Push Insurers Toward Innovative Payment Models Suitable for Era of Personalized Medicine. Personalized Medicine in Brief 2019; 12. 86 Yuste R, Goering S, Arcas BA y, et al. Four ethical priorities for neurotechnologies and AI. Nature 2017; 551: 159–63. 87 Silva GA. A New Frontier: The Convergence of Nanotechnology, Brain Machine Interfaces, and Artificial Intelligence. Frontiers in Neuroscience 2018; 12: 1–8. 88 Towers-Clarck C. Cyborgs Are Here And You’d Better Get Used To It. Forbes. 2018. https://www.forbes.com/sites/charlestowersclark/2018/10/01/cyborgs-are-here-and-youd-better- get-used-to-it/#7600a7b6746a (accessed Oct 15, 2019). 89 DARPA Funds Ambitious Bran-Machine Interface Program. IEEE Spectrum. 2019. https://spectrum.ieee.org/the-human-os/biomedical/bionics/darpa-funds-ambitious-neurotech- program (accessed July 25, 2019). 90 Short Overview of the Human Brain Project. Human Brain Project. 2017. https://www.humanbrainproject.eu/en/about/overview/ (accessed Jan 14, 2020). 91 The BRAIN Initiative. NIH USA. https://braininitiative.nih.gov/about/overview (accessed Jan 14, 2020). 92 Li J, Walker S, Nie J, Zhang X. Experiments that led to the first gene-edited babies: the ethical failings and the urgent need for better governance. Journal of Zhejiang University-SCIENCE B 2019; 20: 32–8. 93 Cyranoski D. Japan approves first human-animal embryo experiments. Nature. 2019. arXiv:2001.09778,2020. 20 https://www.nature.com/articles/d41586-019-02275-3 (accessed July 29, 2019). 94 Kriegman S, Blackiston D, Levin M, Bongard J. A scalable pipeline for designing reconfigurable organisms. Proceedings of the National Academy of Sciences 2020; published online Jan 13. DOI:10.1073/pnas.1910837117. 95 Bourzac K. DARPA Wants Artificial Lifeforms. MIT Technology Review. 2011. https://www.technologyreview.com/s/424325/darpa-wants-artificial-lifeforms/ (accessed Dec 17, 2019). 96 Yetisen AK. Biohacking. Trends in Biotechnology 2018; 36: 744–7. 97 Corea F. Life 3.0 and Biohacking: Rewriting Human Life in the Digital Age. Forbes. 2019. https://www.forbes.com/sites/cognitiveworld/2019/05/14/life-3-0-and-biohacking-rewriting- human-life-in-the-digital-age/#2a63fd056c95 (accessed July 25, 2019). 98 Denton DD. The Weaponizing of Biology: Bioterrorism, Biocrime and Biohacking. Terrorism and Political Violence 2019; 31: 645–6. 99 Bentley PJ, Brundage M, Häggström O, Metzinger T. Should we fear artificial intelligence? European Parliamentary Research Office. 2018. http://www.europarl.europa.eu/RegData/etudes/IDAN/2018/614547/EPRS_IDA(2018)614547_EN.p df (accessed Oct 18, 2019). 100 Weisberg SM, Badgio D, Chatterjee A. A CRISPR New World: Attitudes in the Public toward Innovations in Human Genetic Modification. Frontiers in Public Health 2017; 5: 1–9. 101 Craglia M (Ed), Annoni A, Benczur P, et al. Artificial Intelligence: A European Perspective. Luxembourg: EUR 29425 EN, Publications Office, 2018. 102 High-Level Expert Group on Artificial Intelligence of the European Commission. Ethics guidelines for trustworthy Artificial Intelligence. European Commission. 2018. https://ec.europa.eu/futurium/en/ai-alliance-consultation/guidelines (accessed Dec 12, 2019). 103 European Group on Ethics in Science and New Technologies. http://ec.europa.eu/bepa/europeangroup-ethics/index_en.htm (accessed July 25, 2019). 104 WHO Guideline: recommendations on digital interventions for health system strengthening. World Health Organization. 2019. https://apps.who.int/iris/bitstream/handle/10665/311941/9789241550505-eng.pdf (accessed Oct 23, 2019). 105 Big data and artificial intelligence health ethics. World Health Organization. 2019. https://www.who.int/ethics/topics/big-data-artificial-intelligence/en/ (accessed July 25, 2019). 106 Wiens J, Saria S, Sendak M, et al. Do no harm: a roadmap for responsible machine learning for health care. Nature Medicine 2019; 25: 1337–40. 107 The HUMAINT Project. JRC, European Commission. 2018. https://ec.europa.eu/jrc/communities/en/community/humaint (accessed Jan 1, 2020).