Research Article
Investigating AI in Medical Devices: The Need for Better Establishment of Risk-Assessment and Regulatory Foundations
@INPROCEEDINGS{10.1007/978-3-031-59717-6_4, author={Sandra Baum and Konstantinos Manikas}, title={Investigating AI in Medical Devices: The Need for Better Establishment of Risk-Assessment and Regulatory Foundations}, proceedings={Pervasive Computing Technologies for Healthcare. 17th EAI International Conference, PervasiveHealth 2023, Malm\o{}, Sweden, November 27-29, 2023, Proceedings}, proceedings_a={PERVASIVEHEALTH}, year={2024}, month={6}, keywords={Artificial intelligence Medical device regulation Literature survey Medical device survey}, doi={10.1007/978-3-031-59717-6_4} }- Sandra Baum
Konstantinos Manikas
Year: 2024
Investigating AI in Medical Devices: The Need for Better Establishment of Risk-Assessment and Regulatory Foundations
PERVASIVEHEALTH
Springer
DOI: 10.1007/978-3-031-59717-6_4
Abstract
Artificial intelligence (AI) has the potential to revolutionize healthcare in the EU by addressing challenges, such as shortages of healthcare personnel and more effective diagnosis and care. However, the safety concerns surrounding AI-based medical devices have been a major roadblock to the technology’s wider adoption. This study aims to further investigate these concerns in the European context by analysing the AI-enabled Medical devices currently available in the European Union market along with their potential safety risks. We do this by applying a combination of three research methods: (1) a survey of the safety risks of AI-enabled Medical Devices published between 2012 and 2023, (2) an analysis of AI-based medical devices in the EUDAMED database, and (3) a survey on the perceptions of the EU Medical AI ecosystem stakeholders. Our study analyzed the state-of-the-art with a literature body of 29 papers and summarized a number of risks related to the use of AI in medical devices along with the reported mitigation strategies. Furthermore, we analyzed the approved medical devices (71 devices) that use AI in the EUDAMED database and found that there is a lack of transparency in whether the devices use AI along with the lack of crucial information necessary to asses the devices’ safety risks, such information on training data. Finally, when we survey a number of medical device stakeholders (7 out of 130 respondents) we find that there is a disconnect between the industry and regulators: the medical device representatives emphasize the need for better guidance on post-market surveillance while the regulation representatives feel that they lack expertise in AI.
