CT and Genomics Deep Learning Signature Predicts Cetuximab Sensitivity in Colorectal Liver Metastases

Multi-omics deep learning (CT+genomics) predicted benefit from anti-EGFR therapy in RAS wild-type colorectal liver metastases (AUC 0.86), with HR 17.9 for sensitive cases and better PFS (9 vs 5 months). External validation cohort small.

• Multi-center study with training/testing from prior cetuximab cohort and prospective external validation (enrollment Jan-Dec 2018). Exact cohort sizes not reported in abstract.
• Fusion signature (weighted CT+genomics) achieved AUC 0.86 for cetuximab sensitivity; did not predict chemotherapy response (AUC 0.54).
• Limitation: small external validation cohort; further validation needed.

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