Managing AI/ML Enabled Medical Devices

December 13th, 2023

AI/ML Enabled Medical Devices

Overview

Artificial Intelligence (AI) encompasses the science and engineering behind creating intelligent machines and programs. Techniques in AI range from statistical analysis models to expert systems reliant on if-then statements and Machine Learning (ML). ML, a subset of AI, involves training software algorithms to learn from large sets of data, in this case, medical data and information. These algorithms can be ‘locked,’ maintaining a fixed function, or ‘adaptive,’ altering behaviour based on new data. The FDA is actively involved in evaluating the safety and effectiveness of these AI/ML technologies across various medical applications, recognizing their transformative potential in healthcare.

Trends

The trajectory of AI/ML devices indicates interesting trends. While the year-over-year increase slowed in 2021 (15%) and 2022 (14%) after a significant surge of 39% in 2020 (compared to 2019), projections for 2023 suggest a potential resurgence, with an expected increase of 30% or more based on projected volume.

87% of devices authorized in calendar year 2022 were in Radiology (122), indicating a dominant trend, followed by smaller percentages in other specialties. Through July 2023, Radiology continued to lead with 79% of authorized devices, followed by Cardiovascular, Neurology, Gastroenterology/Urology, Anesthesiology, Ear, Nose and Throat, and Ophthalmic, showcasing a consistent trend in AI/ML device submissions.

The most common uses for AI/ML medical devices include:

Image acquisition and processing
Early disease detection
More accurate diagnosis, prognosis, and risk assessment
Identification of new patterns in human physiology and disease progression
Development of personalized diagnostics
Therapeutics treatment response monitoring

Potential Gaps

AI/ML devices present several challenges and gaps in their development and regulation:

Lack of methods enhancing AI algorithm training for limited labeled training data
Challenges in analyzing and minimizing bias in AI-enabled devices
Need for performance estimation metrics and uncertainty quantification in AI devices
Evaluation methods for continuously learning AI algorithms
Post-market monitoring strategies for AI devices

Regulation

Traditionally, the FDA reviews medical devices through appropriate premarket pathways such as premarket clearance (510(k)), De Novo classification, or premarket approval. However, the FDA’s traditional submission program wasn’t designed for adaptive AI and ML technologies. This led to the development of a proposed regulatory framework for AI/ML-based Software as a Medical Device (SaMD).

The proposed framework includes a “predetermined change control plan” in premarket submissions, envisioning transparency and real-world performance monitoring commitments from manufacturers. This approach aims to enable FDA oversight from premarket development to post-market performance, embracing the iterative improvement power of AI/ML-based SaMD while assuring patient safety.

The FDA is actively working on updating the regulatory framework presented in the AI/ML-based SaMD discussion paper, including issuing draft guidance on the predetermined change control plan.

How can Enerxen help?

Enerxen specializes in supporting companies in regulatory submissions, cybersecurity testing, and best quality assurance practices for AI/ML devices and SaMD systems. Our services aim to assist companies in navigating complex regulatory landscapes and ensuring compliance, facilitating development, and implementing deployment strategies.

Contact us today to get started towards the path of regulatory success with a complimentary consultation with one of our seasoned experts. We will work closely with you to create a personalized strategy that aligns with your product. Find a more in-depth overview to our service what we do here.