Posted by

AI/ML-Based Software for Medical Devices: Artificial Intelligence applications are booming across the industries. Healthcare is one such sector wherein AI applications are all set to revolutionize the way the industry operates. For instance, from drug development to clinical research, the application of AI enables companies to improve patient outcomes with decreased costs and brings in utility, especially in pharma, radiology, and Pathology. As the Artificial Intelligence application in Healthcare Industry is growing, the use of machine learning algorithms in the Medical Device Industry, exclusively for Image analysis, is amplified. It found applications in identifying diseases, effective treatments like personalized medicine, and patients who are on the verge of developing medical conditions.  However, the transformation in healthcare AI is posing new challenges in developing safe and effective medical devices that run on AI algorithm software.

New Framework: In April 2018, the agency granted approvals for AI (Artificial Intelligence) based medical devices which can sense diabetic retinopathy (IDx-DR AI diagnostic system from IDx Technologies Inc. of Coralville, Iowa) and in February 2018, the Viz.AI contact application is a computer-aided triage software that uses an artificial intelligence algorithm to analyze images for pointers connected with a stroke. However, these approved AI algorithms are just “locked” algorithms which don’t always adjust each time the algorithm used. So, these algorithms require a modification, manual verifications as well as substantiation at regular intervals by the manufacturers.  In April 2019, FDA proposed a new regulatory framework in a discussion paper named ‘Proposed Regulatory Framework for Modifications to Artificial Intelligence/Machine Learning (AI/ML)-Based Software as a Medical Device’ for Medical Devices that follows Artificial Intelligence algorithms. This proposal is the first initiation by the FDA to present the artificial Intelligence algorithm operated medical devices to the market. The proposed framework is still in the phase of development which permits the algorithms to progress in ways to improve its performance all through the Medical Device lifecycle. The framework will be followed by the draft guidance with the inputs and feedbacks that FDA will receive. The proposed framework, which is an adaptive algorithm aims in delivering machine learning algorithms which digests the new user data placed in the algorithm from real-world use and offer a diverse yield in comparison to the locked algorithms. For instance, one such system or device would be the one which detects breast cancer abrasions on mammograms and identifies subclasses of cancer too founded on its exposure to real-world data.

Future: Any technological evolution is acceptable in the healthcare industry if it is more powerful, benevolent and valuable. With earlier disease detection, precise diagnosis and more targeted treatments patient population in the Industry will be profited. But, technological changes are strictly unacceptable if it is used unethically, thus harming to a larger population of society. Expecting this proposed framework will provide safe and effective AI/ML software with real-world tracking of the performance once the device is on the market.