A study, published in the European Heart Journal, looked at o study looking at an algorithm to detect coronary artery disease by analysing photographs of a person’s face.
Dr Peter Bannister, Biomedical Engineer & Executive Chair, Institution of Engineering and Technology (IET), said:
“The editorial highlights some fundamental requirements for AI to be applied in a healthcare setting, which this new approach to detecting heart disease does not currently meet: algorithms must first produce measurements which are clinically relevant (the criteria does not necessarily allow patients with an early stage of the disease to be distinguished from healthy patients); they must be trained on a sufficiently large and representative sample of the intended users to be able to perform well across all the variations in patient characteristics such as variation in facial features that you might expect in a real-world population; thirdly the ethical implications – both in terms of patient privacy and potential misuse – need to be considered. Given that the algorithm makes a determination about health based on easily obtainable “selfies” at a stage before the patient themselves might be aware they have the condition, it raises concerns that the algorithm could be maliciously applied on data in the public domain, for example a screenshot from a webconference call being used to discriminate against someone on the basis of their future health insurance premiums.
“That said, if these concerns are properly addressed, there are great benefits from delivering early stage diagnostics using readily-available consumer devices. In general, early detection will enable currently-available treatments which deliver much better outcomes than those which are applied when a disease has already reached a more advanced stage.”
“Feasibility of using deep learning to detect coronary artery disease based on facial photo”, by Lin et al is published in the European Heart Journal.
Editorial: “Selfies in cardiovascular medicine: welcome to a new era of medical diagnostics”, by Kotanidis et al is published in the European Heart Journal.