NIH contract with MRIMATH LLC for an explainable, physician-in-the-loop AI tool to monitor tumors in brain MRIs ($2M) — a federal contract (USAspending)

This contract is one example of medical AI research that the U.S. National Institutes of Health (NIH) funds with public money. For context, patients with brain tumors typically undergo repeated MRI (magnetic resonance imaging) scans so clinicians can watch whether a tumor is regrowing after treatment or whether new lesions appear. Reading these scans is specialized and demanding, and tools to assist that judgment have long been sought. This project sits in that context, aiming to develop software that supports tumor surveillance in brain MRIs.

As for why it matters, the contract foregrounds two explicit design principles. The first is that the system be interpretable, meaning it is built so people can understand why the AI produced a given output. The second is that it be physician-in-the-loop, meaning the AI is designed to support rather than replace the clinician, with the final judgment left to the physician. In medicine, where errors carry serious consequences, this emphasis on explainability and human involvement aligns with long-standing debates in the field about how medical AI should be deployed.

Viewed more broadly, the award shows NIH using a mechanism for backing small, early-stage research and development (purchasing through a solicitation topic) to invest in the theme of imaging-based diagnostic support. Which agencies direct how much money to which companies, and for what purpose, can be traced through open data such as USAspending. Reading a single contract closely makes concrete the kind of technical priorities public funding is steering toward, here AI that supports cancer imaging.