A novel MRI-based model predicts perineural invasion in cholangiocarcinoma

An innovative MRI-based fusion model effectively predicts perineural invasion (PNI) status in intrahepatic cholangiocarcinoma (ICC) patients. The model, incorporating deep learning, radiomics, and clinical features, was developed from a study involving 192 patients. In the training phase, it achieved an AUC of 0.905 and 82.3% accuracy, while the external test set yielded an AUC of 0.760 and 77.8% accuracy. These results indicate its potential to enhance preoperative staging and guide adjuvant therapy decisions.