Deep Learning of Multi-Phase CT Kinetics Maps Intratumoral Heterogeneity to Improve Survival Prediction in Advanced HCC

CT radiomics fusion model improved C-index to 0.713 vs clinical alone 0.651 (P=0.012) in external test; high- vs low-risk median OS 9.5 vs 19.0 months.

• HepaKineticNet generates intratumoral heterogeneity maps by separately encoding arterial wash-in and delayed wash-out kinetics on multiphase CT.
• Risk stratification by the fusion model separated median OS 9.5 vs 19.0 months (HR 3.02, 95% CI 1.86-4.91).
• Retrospective design; prospective validation is needed.

Automated summary

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