AI-PET/CT Radiomics Shows Promise for NSCLC Lymph Node Staging but Needs More Validation

AI-based radiomics on 18F-FDG PET/CT improves distinction between benign and metastatic mediastinal nodes in non-small cell lung cancer. Models combining radiomic features with clinical and conventional PET metrics consistently outperformed radiomics-only approaches. Performance…

• Composite models integrating radiomic features, clinical factors, and deep-learning features consistently outperformed stand-alone radiomics models.
• Performance was lower on external or prospective validation cohorts compared to training or internal datasets, highlighting the need for rigorous prospective multicenter validation.
• Specific approaches designed to address FDG false-positive uptake improved specificity over conventional PET/CT interpretation.

Automated summary

RadPigeon summaries are original and for information only. They are not clinical advice.