Why the Future of Radiology Looks Like GitHub — Not a Doctor

Radiology is hitting a wall. Imaging volumes are growing 5–10% a year. The U.S. faces a projected shortage of up to 30,000 radiologists by 2030. In 2024, UK hospital trusts spent £325 million on temporary radiology workforce solutions — £216 million of that outsourced. Hiring more people isn't a strategy. It's a patch on a structural problem.

Writing for The AI Journal, a thought leader in clinical AI operations argues that the answer lies not in better diagnostics tools, but in rethinking the operating model itself — using the architecture that made complex software scalable: GitHub.

Key Takeaways

• A radiology study is not one task — it's a repository: A single CT scan of the chest, abdomen, and pelvis contains multiple distinct diagnostic problems, each requiring different expertise. The current model assigns all of it to one radiologist. The GitHub model decomposes it: measurements and annotations become commits, subspecialty requests become pull requests, and the signing radiologist becomes the maintainer who holds final accountability.
• AI belongs in the coordination layer, not the diagnosis layer: Most AI investment in radiology is going toward detection and classification — tools that speed up one part of the radiologist's job. The higher-leverage opportunity is coordination: routing tasks to the right expertise, flagging missing pieces, and keeping multi-contributor workflows coherent. That's where AI creates durable value without taking on clinical responsibility.
• Distributed contribution already happens — it's just invisible: Radiologists already seek peer input informally. The GitHub model makes that collaboration structured, time-stamped, attributable, and auditable. Accountability doesn't disappear; it becomes traceable.
• The role shifts from coverage to expertise: In a distributed model, some radiologists become narrow-domain reviewers. Others become final integrators. Some focus on edge cases. The role reorganizes around concentrated expertise rather than blanket ownership of entire studies.

The broader principle extends beyond radiology: any knowledge-work domain that has outgrown the single-author model can benefit from structured, version-controlled contribution architecture.

Read the full article on The AI Journal