Radiology Partners Acquires Cognita to Revolutionize Medical AI

A Strategic Sale for AI in Healthcare

In October 2024, Cognita, a startup specializing in artificial intelligence applied to radiology, made a bold decision by accepting an acquisition offer from Radiology Partners, the largest radiology practice in the world. This decision came less than a year after the company's founding by Louis Blankemeier and his co-founders, who aimed to make their doctoral research useful in the real world.

The AI models developed by Cognita were capable of interpreting complex medical images, such as X-rays and CT scans, for tens of thousands of potential diagnoses. They generated comprehensive radiology reports that reflected how radiologists reason in clinical practice. At a time when AI in radiology was limited to a few specific conditions, this innovation represented a fundamental shift.

Less than a year after its creation, Cognita faced a crucial choice: raise venture capital to continue independently or accept the acquisition offer from Radiology Partners. Conventional wisdom in the tech sector advocates for independence as a sign of true ambition. However, to truly transform healthcare, Cognita chose a different path.

The Challenges of Clinical AI

Clinical AI is a highly regulated field, with long sales cycles and complex dynamics among stakeholders. These constraints make it difficult for a startup to grow independently. By joining Radiology Partners, Cognita hopes to significantly increase access to healthcare on a global scale.

Although Cognita's AI models had demonstrated their effectiveness in research environments, they were not yet ready for large-scale clinical use. Radiology, in particular, presents unique challenges with billions of pixels of data to analyze and critical edge cases to identify.

During his PhD, Louis Blankemeier trained AI models in radiology on substantial datasets, ranging from tens to hundreds of thousands of studies. These models proved their academic value, but in real clinical contexts, they were not yet compliant with the safety and consistency standards necessary for patient care.

Real-world radiology is defined by edge cases where rare but critical pathologies are encountered regularly. A single CT scan can contain 10 high-resolution volumetric series, equivalent to 3D videos. Add prior studies for the same patient, and you can have a billion pixels of data. These billions of pixels encode an amount of information equivalent to entire medical textbooks.

Companies that have made the most progress in AI control the entire system and scale rapidly. They own the vehicles, the sensor stack, the data collection pipeline, the simulation environments, and the deployment infrastructure. This integration, operating at scale, allows them to continuously collect rare edge cases, retrain models, validate improvements, and redeploy safely.

Integration and Scale: Keys to Success

Successful companies in the AI field control the entire system, from data collection to deployment infrastructure. This integration allows for the continuous collection of rare edge cases, retraining of models, and validation of improvements. In radiology, this requires diverse historical datasets and live data streams.

Cutting-edge language models have shown that continuous human feedback is essential for improving AI models. In radiology, every report generated by AI must be reviewed and approved by a human radiologist, creating a virtuous cycle of continuous improvement.

Access to this correction data is rare and can only function meaningfully at a massive scale. These capabilities would be incredibly difficult to achieve as an independent AI startup.

Moreover, cutting-edge language models have clearly demonstrated that continuous, high-quality, and extensive human feedback is key to making models useful. This is equally true in radiology. In a world where radiology reports are drafted by AI, every draft must be examined, modified, and approved by a human radiologist. These modifications become high-quality signals that can be leveraged to improve AI models. Better models enhance the accuracy and capacity of radiologists. Increased accuracy among radiologists improves the quality of future training data. Enhanced capacity allows radiologists to take on additional contracts. This, in turn, generates more high-quality data and corrections, setting in motion a powerful virtuous cycle.

Credibility Through Integration

In the healthcare sector, trust is built on rigorous and real evidence. For a new technology to be adopted, it must demonstrate its clinical effectiveness, reliability, and safety. By integrating with Radiology Partners, Cognita hopes to prove its effectiveness at scale and establish its credibility in the field.

Evidence in healthcare is not generated by small pilots. It is built through sustained performance across diverse sites, patient populations, modalities, and edge cases. If a system proves its effectiveness within the largest radiology practice in the world, it establishes its credibility across multiple dimensions: effectiveness, reliability, safety, and scalability.

Louis Blankemeier, CEO and co-founder of Cognita, has always been driven by the mission to increase access to healthcare through technology. Convinced that AI is the most promising technology to achieve this, he pursued a PhD in AI at Stanford, where he developed Merlin, a 3D vision-language model for interpreting CT scans, published in "Nature" in 2026.

Cognita is now the AI business unit of Mosaic Clinical Technologies at Radiology Partners. This strategic integration aims to enhance the efficiency and reach of radiological diagnostics while accelerating global access to healthcare.