Background: Artificial Intelligence (AI) has been evolving in the field of medical devices. However, robust and concrete clinical data are still required to achieve optimum efficacy and safety with the use of medical device.
Objectives: 1) Advancement of artificial intelligence in medical device and software. 2) Risk vs benefit profile of medical devices in diverse therapeutic area. 3) Need of mitigation strategy to safeguard the use of medical device.
Methods: In this study we have performed screening of AI software products on (www.aiforradiology.com). Further pubmed, scopus, EUDAMED and US FDA database were screened to collect the various efficacy, feasibility and reliability data of respective medical device to assess their risk vs benefit profile. We have randomly collected the data of 4 recent breakthroughs in AI technologies and their therapeutic or diagnostic applications. The AI implementation data was collected for various therapeutic field viz. Diabetes (IDx-DR: Diabetic Retinopathy, Guardian connect: Glucose monitoring), Orthopedic (Osteodetect), CNS (Embrace2: seizure monitoring device), CVS (Fibricheck). The clinical data and accuracy of above mentioned devices have been gathered from available sources and risk: benefit profile reviewed. The vendors for above mentioned 4 medical devices were contacted independently to validate the collected information. We retrieved information regarding their clearance from concerned regulatory authorities. The CE status of respective 4 medical device was screened. Moreover, the FDA approval status was collected and confirmed with publicly available database.
Results: The overview consisted 110 CE-marked AI products. For 71/110 products, no peer-reviewed evidence of efficacy were found. We found diverse heterogeneity in cost, deployment methods, and regulatory classes. Out of 4 of detailed reviewed devices, we found that major gap in usability is found due to data credibility recorded through devices. Further, cost and availability play a major role in popularity of these devices. Since, these technologies are evolving a secondary medical professional's opinion is required for final diagnosis. which, further limits their usage.
Conclusions: Artificial intelligence in medical device is advancing on a considerable scale. However, robust data accuracy, re-assessment of methods, validity need to be kept on check for strengthening the credibility of AI in medical devices.
