Using Claude Code for Second-Opinion Medical Image Analysis
A developer used Claude Code with Opus 4.8 to analyze MRI DICOM files, while simultaneously using GPT 5.5 Pro to audit clinical treatment protocols. The process highlights the potential of AI tools to flag discrepancies in medical care plans.
Why it matters
Demonstrates the practical, albeit experimental, use of autonomous agents in auditing medical procedures and reviewing technical diagnostic data.
TL;DR
- 01AI can act as an auxiliary tool for second opinions, but results vary significantly based on prompt context.
- 02Separating tasks—using different models for auditing protocols versus image processing—improved clarity.
- 03Results should never replace professional medical advice.
Setup and Execution
The analysis was performed using Claude Code with Opus 4.8 (xhigh) for image analysis, while GPT 5.5 Pro handled protocol auditing. The dataset consisted of several hundred DICOM files without extensions (266 MB).
Methodology
1. Data Prep: The agent was instructed to install necessary Python dependencies for DICOM processing. 2. Analysis: Claude Code was given the medical context of shoulder pain and tasked with creating a detailed investigation plan. 3. Auditing: The user compared the physician's report against clinical guidelines using GPT 5.5 Pro, flagging treatments like shockwave therapy and homeopathic injections.
Limitations
The model provided inconsistent results, initially identifying an intact tendon where a tear was suspected, requiring a second, more context-rich prompt to reach a refined conclusion.