Computer vision models (CVMs) surpass large language models (LLMs) in accurately labeling incidental breast abnormalities on chest CT scans.
Key Details
- 1Study presented at 2024 SIIM, led by University of Wisconsin researchers.
- 2Compared Covera Health’s CVM to Qwen2 7B LLM on 17,752 chest CTs in women aged 40-72 (2015–2017).
- 3Incidental breast abnormalities present in 1–7% of chest CTs; 30% of these cases are malignant.
- 4CVM had 97.5% accuracy and 81.9% sensitivity vs. LLM’s 95.7% accuracy and 39.8% sensitivity.
- 5CVM showed higher F1 score (0.81) vs. LLM (0.54); LLM had marginally higher PPV (82% vs. 79.7%).
- 6Researchers suggest continued 'AI vs AI' testing with human adjudication could aid in real-world validation.
Why It Matters
Automated detection of incidental breast findings on chest CT could reduce missed malignancies and streamline workflow. The superior sensitivity of CVMs—validated on a large real-world dataset—underscores their potential for clinical integration in radiology AI workflows.
Source
AuntMinnie
Related News
•Radiology Business
Meta Funds Ohio State’s Radiology AI Innovation Projects
Meta will fund Ohio State researchers to explore AI-driven advancements in radiology, including chest X-ray reporting and disease prediction.
•AuntMinnie
ChatGPT-5 Performs Poorly at Detecting Distal Radius Fractures on X-rays
ChatGPT-5 missed about half of distal radius fractures on x-rays, showing limited diagnostic capability in a real-world scenario.
•AuntMinnie
AI Triage Tool Speeds Up ICH Head CT Turnaround in ED
AI-based triage for intracranial hemorrhage on head CT reduced emergency department report turnaround by 13 minutes.