A Multichannel and Multimodal Combined Model Based on 2.5D Deep Learning For Predicting Tumor-Free Survival After Liver Transplantation In Hepatocellular Carcinoma: A 7-Year Retrospective Study.

Authors

Hao Xie, Jianli Yang, Yong Luo, Ya Yang, Qiongling Hu, Shanchuan Liao, Bin Shi, Junzhen Fan, Han Meng, Mengyi Wang, Xuetao Mu

Published in

Academic radiology. Nov 10, 2026. Epub Nov 10, 2026.

Abstract

Hepatocellular carcinoma (HCC) accounts for 90% of primary liver cancers and ranks as the fourth leading cause of cancer-related deaths worldwide. This study aimed to promptly identify high-risk patients for early recurrence after liver transplantation (LT), thereby optimizing personalized follow-up and intervention strategies for patients with HCC.
A retrospective cohort of 143 patients with HCC was selected. Following the collection of imaging investigations, clinical data, and postoperative follow-up information, we developed a combined model based on 2.5-dimensional (2.5D) deep learning (DL). This model integrates clinical features, radiomics characteristics, and DL data across multiple modalities, utilizing multichannel 2.5D data comprising coronal, sagittal, and peritumoral slices of the largest tumor area. Model performance was assessed through receiver operating characteristic (ROC) curves, calibration curves, and decision curve analysis (DCA).
The 2.5D DL model demonstrated superior performance over the 2D DL model, achieving area under the ROC curve (AUC) values of 0.784/0.842, 0.799/0.843, and 0.824/0.827 in the training and testing cohorts, compared to the 2D model's 0.736/0.819, 0.751/0.818, and 0.724/0.803. The combined model significantly surpassed other models, attaining C-index values of 0.789/0.826 in the training and testing cohorts, along with 1-year, 2-year, and 3-year AUC values of 0.842/0.931, 0.865/0.891, and 0.853/0.885, respectively. Furthermore, the combined model demonstrated the highest consistency in calibration curves and provided the greatest clinical net benefit in DCA.
The combined model based on 2.5D DL demonstrates robust applicability and high predictive performance in identifying early postoperative recurrence of HCC. This strategy facilitates proactive and individualized intervention for high-risk patients, thereby improving clinical outcomes.

PMID:
43009362
Bibliographic data and abstract were imported from PubMed on 11 Nov 2026.

Read full publication at:
Please sign in to see all details.

Stats

1-terrible, 9-excellent. How would you rate this publication? Sign in in to submit your rating.

Post a comment

You need to be signed in to post comments. You can sign in here.

Comments

There are no comments yet.