Radiology Partners ramped up its investment in AI by acquiring Cognita Imaging, a startup that’s developed AI vision language models for analyzing CT and X-ray images and drafting initial radiology reports. RP executives see the acquisition as going beyond traditional point-source AI models and toward a future where AI automates much of the traditional image interpretation process.
The $80M acquisition expands on an equity stake RP already had in Cognita, which had been operating in stealth mode since its spin-off from Stanford University’s Center for Artificial Intelligence in Medicine and Imaging lab.
- Cognita was formed by a team led by CEO Louis Blankemeier, PhD, to commercialize Stanford research on vision language models, a type of generative AI that’s far more versatile than the traditional point-source models being commercialized to analyze medical images.
Instead, Cognita’s technology is able to analyze text as well as CT or X-ray images and produce first drafts of radiology reports that just need a radiologist’s review and signature to be complete.
- Extremely positive clinical tests with Cognita’s VLM models spurred RP to acquire the rest of the company it didn’t already own, said Rich Whitney, chairman and CEO of Radiology Partners.
Cognita’s technology powers Mosaic Drafting, RP’s new application for helping radiologists draft reports that operates under the company’s recently launched Mosaic Clinical Technologies branding. Early clinical testing has found that Mosaic Drafting…
- Increases radiologist detection rates by 52%.
- Results in a fourfold decline in radiologist errors.
- Reduces radiologist reading times by up to 76%.
RP plans to deploy Mosaic Drafting through Mosaic Clinical Technologies, which the company launched in July as the technological foundation for a massive rollout of AI across its physician practices.
- Mosaic Chief Medical AI Officer Nina Kottler, MD, said Mosaic Drafting is currently being used within Radiology Partners under IRB approval, but the company will pursue an FDA authorization – most likely under a de novo pathway – that probably will come sometime in 2026.
In a broader sense, RP sees Mosaic Drafting and other VLM tools as key to the growing mismatch between rising imaging volume and stagnant radiologist supply – a mismatch that can only be solved through greater automation.
- And as the largest private radiology organization in the U.S., Radiology Partners has the organizational heft to make VLMs work on a wide scale.
The Takeaway
RP’s acquisition of Cognita is a major development in putting vision language models on the fast track to real-world clinical use. Unlike point-source AI, VLMs could hold the key to really solving radiology’s volume overload dilemma.
