Which Generative AI Model is the Best Diagnostician?

0

1. Study evaluates LMMs for lung cancer CT scan diagnostics.

2. Models tested include GPT-4V, LLaVA-1.5, and BiomedCLIP.

3. GPT-4V had the highest agreement of 75 percent.

4. Accuracy for NCCN risk assessment was under 50 percent.

5. LLaVA-1.5 produced the most consistent descriptive results.

6. Models exhibited hallucinations and alignment issues.

7. Future improvements needed through collaboration with radiology professionals.