DCE-MRI habitat fusion model accurately predicts microvascular invasion in liver cancer

A transformer-based model combining DCE-MRI habitat and conventional radiomics with clinical predictors achieved an AUC of 0.923 (95% CI 0.858-0.988) for predicting microvascular invasion in hepatocellular carcinoma, and stratified recurrence-free survival (p=0.033).

• A fusion model integrating habitat radiomics, conventional radiomics, and clinical features (pseudocapsule integrity, tumor diameter) outperformed transformer-based single-modality models for MVI prediction.
• Model achieved AUC 0.950 (95% CI 0.918-0.981) on training data and 0.923 on testing data with robust calibration.
• Study included 193 patients from two centers, but an external validation cohort was not reported.

Automated summary

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