Radiology Partners petitions FDA for greater clarity around imaging AI regulations

Radiology Partners and its technology services division are petitioning the U.S. Food and Drug Administration, seeking greater clarity around the regulation of vision language models. 

Mosaic Clinical Technologies first submitted the citizen petition to the agency on Aug. 12. It’s asking the FDA to take action to clarify the regulatory status of commercially available AI models—which perform medical image analysis to support diagnosis and treatment. 

Rad Partners and Mosaic believe diagnostic imaging vision-language models represent an “important and promising” advancement in medicine. However, they are asking for regulatory clarity regarding the standards that should apply when such technologies are distributed for diagnostic use. 

“As these technologies are increasingly commercialized and deployed in clinical settings, differing interpretations have emerged regarding the application of existing FDA medical device requirements, creating uncertainty for developers, healthcare organizations, clinicians and patients,” Mike Peresie, president of Mosaic Clinical Technologies, wrote to the FDA. “Clarity on standards will ultimately accelerate both the development and adoption of safe and trusted clinical diagnostic technologies.”

Peresie said the petition is seeking to clarify how existing regulatory principles should be applied to commercially distributed imaging vision-language models. More specifically, Rad Partners wants to know whether a model marketed with the expectation of subsequent fine-tuning on institution-specific data is itself a medical device requiring premarket clearance or approval. The radiology group also wants the FDA to take other actions in its authority to ensure consistent application of medical device requirements to available vision language models.

Experts at Rad Partners believe the growing prevalence of these foundation models—trained on large-scale datasets to solve a variety of downstream tasks—raises questions regarding their safety and effectiveness. Currently, there is no established standard for validating the performance and quality of these AI models. 

“The absence of regulatory oversight and standardization significantly increases the risk that they may be trained on biased or unrepresentative datasets and that they lack transparency, explainability, and basic cybersecurity controls,” Peresie wrote. “All these risks have the potential to cause tangible patient harm. In addition, several developers disclaim responsibility for the models’ reliability or compliance with various legal requirements and provide few, if any, warranties or representations regarding the performance, reliability, or fitness of the models for the various diagnostic and clinical uses for which they are marketed.”

As a result, downstream users such as radiologist bear “substantial responsibility” for determining whether these models are appropriate for their intended use. Models introduced without premarket review may not have demonstrated the baseline safety and effectiveness required for FDA authorization, he added.  

Rad Partners is concerned that conflicting interpretations of regulatory requirements could create inconsistent compliance. It wants the FDA to clarify all stakeholders’ responsibilities in the deployment of diagnostic foundation models. Clear expectations around validation, performance monitoring, transparency and quality management “would support patient safety, promote responsible innovation and provide greater consistency across the healthcare ecosystem.” 

“For the aforementioned reasons, we respectfully request that FDA grant the actions requested in this petition,” the letter concludes. “Clarification from the agency would benefit patients, healthcare providers, developers, healthcare organizations, researchers, and other stakeholders by promoting consistent interpretation of existing medical device requirements while continuing to support responsible innovation. We appreciate the FDA’s leadership in protecting public health and supporting safe innovation in medical imaging and AI.”