Here’s a summary of the article,focusing on the key findings about ChatGPT-5‘s performance in breast cancer detection:
ChatGPT-5 Shows Enhancement,But Still Trails Experts & Specialized AI
A recent study evaluated ChatGPT-5’s ability to identify breast cancer in medical imaging. While ChatGPT-5 outperformed previous versions, it still substantially lagged behind both human radiologists and AI specifically designed for breast cancer screening.
Key Findings:
CBIS-DDSM Dataset:
Cancer Identification Accuracy: 63.5% (Doctors: 86.9%)
Normal Tissue Classification Accuracy: 52.3% (Doctors: 88.9%)
Embed Dataset: ChatGPT-5 showed its best results within the ChatGPT series, with varying accuracies across different tasks (density, deformation, categories, calcification, cancer judgment) ranging from 52.5% to 64.5%.
* Comparison to Specialized AI (MAMMO-CLIP): MAMMO-CLIP achieved 82.3% accuracy in cancer discrimination,almost 30% higher than ChatGPT-5’s performance.
Conclusion:
The study highlights that while general-purpose AI like ChatGPT-5 is improving, it still struggles to match the precision of AI models specifically trained for medical applications like breast cancer screening. This underscores the importance of specialized AI in healthcare for accurate diagnoses.
