Back to all news
Large Language Models Rival Physicians in Complex Lung Cancer Decisions
Tags:Research
A real-world study reveals that large language models (LLMs) can match or exceed human physicians' performance in challenging lung cancer case decision-making, especially for rare cases.
Key Details
- 150 challenging lung cancer cases (complex, rare, refractory) were evaluated using blinded, multidimensional scoring by experts.
- 2LLMs reviewed: DeepSeek R1, Claude 3.5, Gemini 1.5, and GPT-4o; physician decisions stratified by experience; some juniors received AI assistance.
- 3DeepSeek R1 performed between intermediate and senior physicians overall; LLMs outperformed intermediates in rare cases but lagged in refractory (longitudinal) cases.
- 4AI-augmented junior physicians saw 80-90% boosts in comprehensiveness and specificity for rare cases, but specificity slightly dropped for refractory cases.
- 5Error profiling showed LLMs are strong in knowledge breadth/updates, while physicians excel in longitudinal reasoning and stability.
Why It Matters
This study highlights when and how LLMs can best augment clinical decision-making—showing strong synergies for rare cases and pointing to limitations in complex, longitudinal care settings. These insights will inform the integration of AI tools into radiology and oncology multidisciplinary teams for case-based decision support.
Source
EurekAlert
Related News
•EurekAlert
AI Tool Predicts Risks for Cancer Patients After Heart Attack
A new AI-powered risk prediction model assists clinicians in treating cancer patients following a heart attack by combining oncology and cardiovascular data.
•EurekAlert
AI Reveals Abdominal Muscle Density as Predictor of Fall Risk in Middle Age
Mayo Clinic researchers demonstrate that AI analysis of abdominal CT scans can help predict fall risk in adults as early as middle age.
•EurekAlert
Natural-Looking Adversarial Images Could Strengthen Medical AI Security
Japanese researchers introduce IFAP, a new technique that generates natural-looking adversarial images to more effectively test and improve AI vision systems.