In recent years, a burgeoning interest in and concern over the use of artificial intelligence (AI) in medicine and healthcare has stood at the centre of interdisciplinary scientific research, political debate, and social activism. The goal of this report is to explain the areas in which AI can contribute to the medical and healthcare field, pinpoint the most significant risks relating to its application in this high-stakes and quickly-changing field, and present policy options to counteract these risks, in order to optimise the use of biomedical AI. Not only will this ensure the safety and respectful treatment of patients receiving AI-mediated healthcare, it should also aid the clinicians and developers involved in implementing it.
This study employs an interdisciplinary methodology based on a comprehensive (but nonsystematic) literature review and analysis of existing scientific articles, white papers, recent guidelines and regulations, governance proposals, AI studies, and online publications. The multidisciplinary resources examined for this report include works from the fields of computer science, biomedical research, the social sciences, biomedical ethics, law, industry, and government reporting. This report explores a wide range of technical obstacles and solutions, clinical studies and results, as well as government proposals and consensus guidelines.
This study first outlines the potential for AI in medicine to address pressing issues, in particular the ageing population and the rise of chronic diseases, a lack of health personnel, inefficiency of health systems, lack of sustainability, and health inequities. The report also details the different fields in which biomedical AI could make the most significant contributions: 1) clinical practice, 2) biomedical research, 3) public health, and 4) health administration.
In the realm of clinical practice, the report goes into further detail concerning specific contributions – both realised and potential – to particular medical areas such as radiology, cardiology, digital pathology, emergency medicine, surgery, medical risk and disease prediction, adaptive interventions home care, and mental health. In biomedical research, the report details the potential contributions of AI to clinical research, drug discovery, clinical trials, and personalised medicine.
Lastly, the report presents potential contributions of AI at the public health level as well as to global health.
This study identified and clarifies seven main risks of AI in medicine and healthcare: 1) patient harm due to AI errors, 2) the misuse of medical AI tools, 3) bias in AI and the perpetuation of existing inequities, 4) lack of transparency, 5) privacy and security issues, 6) gaps in accountability, and 7) obstacles in implementation. Each section, as summarised below, not only describes the risk at hand, but also proposes potential mitigation measures.