Epsilon Health has raised a $20 million Series A, Axios reports. The company combines a radiology practice with internally developed AI, putting software into the work of physicians who interpret medical images. Its approach is to improve the path from a scan arriving to a completed clinical report, rather than build a stand-alone chatbot for patients.
The connection between an image and its report is central. On technology supplier Sirona's site, Epsilon founder Rustin Rassoli describes building the company's AI where models, images and reports can work together. Sirona's platform links findings in a report to the corresponding anatomy in an image and can assist with drafting an impression—the radiologist's summary of what a scan shows. Those are capabilities of the underlying supplier platform, not proof that every feature is part of Epsilon's own model.
Epsilon is a customer of that infrastructure, not its owner. Rassoli says in a separate Sirona customer account that Epsilon's physicians use custom AI workflows and still make the final clinical call. That distinction explains the operating proposition: software assists a working medical practice, and useful output has to fit how its radiologists read and report cases. It does not establish that an algorithm can independently diagnose a patient or replace the reviewing physician.
The founder ties the business to seeing his father's imaging center contend with growing backlogs and radiologist shortages. In a public post, he argues that recognizing a finding in an image is only one part of the job; a practical system must address the process through to the final report. He describes bringing radiologists, researchers and engineers together inside an integrated practice. This is the company's rationale for its structure, not an independent finding that the wider shortage has been solved.
The test is whether that close connection between development and daily clinical work produces dependable gains as the practice expands. Faster drafts are useful only when physicians can check them efficiently and preserve the quality of the completed report. Epsilon's emphasis on retaining the radiologist's decision makes that review step a central part of the business, rather than a task to be removed.