Get a second read of your own MRI scan files
Antoine Finkelstein had Claude Code analyze his shoulder MRI files; it argued, with moderate-to-high confidence, for no tear where the radiologist reported one, and radiologists pushed back.
Done with Claude Code
What they did
Developer Antoine Finkelstein gave Claude Code the 266MB of raw scan files from his own shoulder MRI. He had it write and run its own analysis programs. Then it compared its reading with the radiologist's report. He chose what to ask and how far to trust the answer. No doctor checked the results.
How it went
The agent concluded with "moderate-to-high confidence" that the evidence pointed to no clear tear. The radiologist had reported a Grade III partial-thickness tear.
Worth knowing
Radiologists on Hacker News pushed back on the agent's reading. Treat an answer like this as a question to bring to a doctor, not a diagnosis.
Try it yourself with Claude Code
I have the DICOM files from my [body part] MRI in [folder path] and the radiologist's report in [report file]. Inventory the files and explain which series and views are present. Write and run scripts to view and analyze the images relevant to the report's findings. Then compare your observations with each finding in the report. For each one, say whether you agree, what evidence you see, and how confident you are. Clearly list the limits of this analysis. This is not a diagnosis. Give me a list of questions to ask my doctor.
Swap the highlighted [parts] for your own details.
Source: HotHardware
Written with AI assistance from the original post and checked against our editorial guidelines. Results are as reported by the source, not independently verified.
