I used Claude Code to get a second opinion on my MRI

4 hours ago

The author experienced right shoulder pain and received an MRI diagnosis of a Grade III partial-thickness tear at the subscapularis tendon insertion, with immediate treatment suggestions that felt premature.
Upon reviewing the MRI results and treatment with GPT 5.5 Pro, concerns were raised about the use of shockwave therapy (against guidelines) and a homeopathic injection (Traumeel) lacking therapeutic indication.
Using Opus 4.8 (via Claude Code) to analyze the MRI DICOM files, the AI reported an intact tendon, contradicting the human doctor's finding of a significant tear.
To arbitrate, the author provided Opus with both the human report and additional context from ChatGPT, leading to a methodical review that concluded mild tendinosis with no tear, favoring the AI's analysis with moderate-to-high confidence.
The discrepancy left the author in uncertainty about trusting AI versus human diagnosis, highlighting AI's potential for second opinions but also current limitations in medical trust and decision-making.

Hasty Briefsbeta