Clinical usability of an explainable AI decision support tool and evaluation of multimodal models in NSCLC.

AI models using clinical and blood data achieved up to 0.77 AUC in NSCLC treatment prediction, outperforming traditional clinical scores in the largest real-world study to date.

• Why it matters: Current immunotherapy selection in NSCLC relies on imperfect biomarkers, creating a need for more accurate, reliable decision support tools to improve patient outcomes.
• What they did: The study integrated multimodal data—clinical, blood, imaging, pathology, and genomics—into machine learning and deep learning models, evaluating their performance and clinical usability across 2,396 patients.
• The result: AI models significantly outperformed traditional scores, and clinicians improved predictions with explainable AI tools, supporting the clinical adoption of AI-based decision support in NSCLC.