Cognita Imaging wins FDA contract to test LLMs in evaluating radiology AI

Cognita Imaging has been awarded a contract by the U.S. Food and Drug Administration to test a method for evaluating large language models used in assessing radiology AI, according to MedTech Dive.

The arrangement centers on Cognita’s approach to scoring AI-generated radiology reports, work the outlet frames as part of the FDA’s broader effort to determine how it will regulate generative AI-enabled medical devices. MedTech Dive reports that the contract gives the agency a concrete testbed for evaluation methods at a moment when generative AI tools are increasingly proposed for use in diagnostic imaging workflows.

Radiology has been an early proving ground for medical AI, with software already cleared for tasks such as flagging abnormalities on scans. Generative AI, which can draft narrative reports rather than simply detect features, poses different evaluation challenges for regulators, since assessing the accuracy and reliability of generated text differs from assessing detection performance.

The contract’s specifics, including its scope, timeline, and evaluation criteria, are not detailed in the available reporting. MedTech Dive characterizes the award as a step by the FDA toward developing methodology it could apply to future submissions involving generative AI in medical devices, without specifying further regulatory decisions that may follow.

No additional detail on contract value, duration, or next steps is available in the source material.