Meta-analysis: AI Radiomics for PD-L1 Prediction in NSCLC Shows Moderate Accuracy

Meta-analysis: AI radiomics models predict PD-L1 expression in NSCLC with AUC 0.78 (TPS≥1%) and 0.80 (TPS≥50%); ML better at TPS50, DL at TPS1.

• Systematic review and meta-analysis of 35 diagnostic-accuracy studies evaluating AI radiomics models for PD-L1 status in non-small cell lung cancer.
• Subgroup analyses: machine learning outperformed deep learning for predicting TPS≥50%, while deep learning was superior for TPS≥1%. Imaging modality also influenced accuracy, but findings varied by TPS threshold.
• All included studies were retrospective, and none reported external validation; results should be interpreted cautiously due to potential overfitting and heterogeneity.

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