Cognita Imaging Receives $1.29 Million FDA Contract to Test New Approach to Evaluating Generative AI in Radiology

Research will test whether multiple language models can work as a jury to assess AI-generated reports across approximately 1 million patient exams, with radiologists reviewing clinically important discrepancies

NASHVILLE, Tenn. & PALO ALTO, Calif.--(BUSINESS WIRE)--Cognita Imaging, Inc. (Cognita), a wholly owned subsidiary of Mosaic Clinical Technologies, Inc. (Mosaic), has received a $1.29 million research contract from the U.S. Food and Drug Administration (FDA) to test a new way of evaluating AI-generated radiology reports. The contract took effect June 22, 2026, and will run for 18 months.

The project, titled “Virtual Subject Matter Expert Agents for Radiology Report Evaluation,” focuses on a growing challenge in medical AI, which is how to rigorously evaluate systems that generate complete radiology reports. Traditional reader studies are essential, but expert review takes time and is difficult to extend to the number and variety of cases needed to understand how a model may behave in day-to-day practice.

Under the contract, Cognita will develop and validate a framework that uses several large language models (LLMs) to evaluate human-drafted and AI-generated radiology reports. The work will be led by Akshay Chaudhari, Ph.D., Cognita co-founder, associate professor of radiology and biomedical data science at Stanford University and principal investigator on the contract. Louis Blankemeier, Ph.D., co-founder and chief executive officer of Cognita, is the co-investigator.

“Once an AI system starts writing the entire report, the evaluation problem changes,” said Chaudhari. “A few hundred cases may tell you whether a model works in a narrow setting. They do not show every way it can fail in practice. We want to test whether a group of language models, with radiologists reviewing the difficult cases, can give us a clearer and more repeatable picture at real-world scale.”

Cognita calls the method “LLMs-as-a-jury.” Rather than asking one language model to grade another model’s report, the framework will compare the judgments of several models. The goal is to make the evaluation more consistent, reduce dependence on the tendencies of any one model and identify where automated review can meaningfully extend expert evaluation.

After building and testing the framework, Cognita will apply it to approximately 1 million patient exams from a large and diverse U.S. cohort. Researchers will examine performance across different patient groups, care settings, imaging equipment and diseases, including uncommon findings that may appear too rarely to be studied well in a smaller dataset. The team will also recreate smaller validation cohorts from the larger dataset to determine what information may be missed when a study includes fewer cases.

Radiologists will review clinically important disagreements and determine whether the problem came from the AI-generated report, the language-model jury or the original radiologist’s report. Cognita will use those findings to improve the framework and study how automated review and physician expertise can work together in both premarket validation and post-market monitoring.

“Radiologists know the edge cases matter,” said Nina Kottler, M.D., chief medical AI officer at Mosaic Clinical Technologies. “A model can look strong in a carefully selected study and still struggle at another hospital, on different equipment or with a rare presentation. This work lets us look for those differences at a scale that would be very difficult to reach through reader studies alone, while keeping radiologists involved where their judgment matters most.”

The research builds on Cognita’s prior work in radiology-specific language and vision-language models. That work includes GREEN, an open-source tool that compares reference and AI-generated radiology reports and identifies clinically meaningful differences. Cognita will also draw on clinical expertise, technology infrastructure and real-world radiology workflows from the broader Mosaic and Radiology Partners Clinical Services’ ecosystem.

During the contract, Cognita will provide FDA with software code and practical guidance for building LLM juries, analyses comparing large and small validation cohorts, radiologist-reviewed discrepancy studies, and reports on the project’s findings and limitations. The contract was awarded through FDA’s Broad Agency Announcement program for advanced research and development in regulatory science.

The contract supports research activities related to regulatory science and does not represent FDA endorsement, approval, or clearance of any Cognita technology or product.

About Cognita Imaging, Inc.

Cognita Imaging, Inc. develops vision-language models and other generative AI capabilities for radiology. Its work is focused on helping radiologists deliver accurate, efficient care and expanding access to high-quality health care. Cognita seeks to elevate the performance of all radiologists to the level of the best expert subspecialists and, longer term, to empower radiologists with digital imaging information not easily perceptible to the human eye. Cognita is a wholly owned subsidiary of Mosaic Clinical Technologies, Inc. Learn more at cognita.ai.

About Mosaic Clinical Technologies, Inc.

Mosaic Clinical Technologies, Inc., a wholly owned subsidiary of Radiology Partners, Inc., develops technology for medical imaging. Its MosaicOS™ platform brings diagnostic tools, AI and radiology workflows together in a cloud-native operating system designed to support radiologists and the organizations they serve. Learn more at MosaicClinical.ai.

About Radiology Partners, Inc.

Radiology Partners, Inc., through RP Clinical Services and its managed practices, together with Mosaic Clinical Technologies, Inc., provides radiology services and technology to more than 3,400 hospitals and other health care facilities. Its network includes more than 4,000 radiologists who interpret more than 55 million imaging studies each year. Learn more at radpartners.com.