Predicting microvascular invasion in hepatocellular carcinoma using contrast-enhanced CT-based deep learning models: a comparative study of transformer, CNN, and MIL
Abstract
ObjectiveMicrovascular invasion (MVI) predicts postoperative recurrence in hepatocellular carcinoma (HCC), yet preoperative prediction remains challenging. This study compared a Transformer, a convolutional neural network (CNN), and a multiple instance learning (MIL) framework for predicting MVI from contrast-enhanced CT.MethodsWe retrospectively enrolled 236 patients with pathologically confirmed HCC. Tumor regions of interest were semi-automatically delineated on arterial-phase CT images, and slices centered on the largest tumor cross-section were sampled to build a 2.5D dataset. Four models—Transformer, MIL, CNN, and a clinical baseline—were trained. Models were compared using the area under the receiver operating characteristic curve (AUC) with the DeLong test; bootstrap resampling assessed the stability of AUC estimates.ResultsThe Transformer achieved the highest AUC of 0.827 (95% CI 0.7585–0.8962), with an accuracy of 0.775, sensitivity of 0.577, and specificity of 0.889, followed by MIL (0.766; 95% CI 0.6875–0.8441), CNN (0.723), and the clinical model (0.715). Bootstrap validation confirmed stable AUC estimates for all models, with the Transformer ranking first (mean 0.827; SD 0.034). The Transformer significantly outperformed the CNN (adjusted P = 0.001) and the clinical model (adjusted P = 0.039); after correction for multiple comparisons, its difference from MIL was of borderline significance (adjusted P = 0.097). No significant differences were observed among the CNN, MIL, and clinical models.ConclusionAmong the four models evaluated, the contrast-enhanced CT-based Transformer predicted MVI most accurately, and its advantage persisted after bootstrap internal validation. These findings concern the ranking of discriminative performance only; the clinical utility of the model is not established, and prospective validation in larger cohorts is required before any clinical application.
Related
- Imaging biomarkers related to 18F-FDG PET/CT for predicting microvascular invasion in hepatocellular carcinoma: a systematic review and meta-analysis
- Microvascular invasion-associated serum glutaminolysis index and recurrence risk after liver transplantation for hepatocellular carcinoma
- Magnetic resonance imaging-based radiomics for predicting P53-mutated hepatocellular carcinoma: intratumoral and peritumoral analysis
- Construction of a five-gene prognostic model involving T cell exhaustion and metabolic reprogramming for predicting hepatocellular carcinoma risk using LASSO: insights into immune infiltration and drug sensitivity
- Interaction between microvascular invasion and early recurrence of hepatocellular carcinoma in cirrhosis patients with curative liver resection: a retrospective cohort study.
BibTeX
@article{lin2026predicting,
title = {{Predicting microvascular invasion in hepatocellular carcinoma using contrast-enhanced CT-based deep learning models: a comparative study of transformer, CNN, and MIL}},
author = {Mianlei Lin and Xiaojun Zheng and Caitao Zhao and Tao Lin and Qishu Zhao},
journal = {Frontiers in Oncology},
year = {2026},
month = nov,
doi = {10.3389/fonc.2026.1988882},
url = {https://www.frontiersin.org/articles/10.3389/fonc.2026.1988882},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Predicting microvascular invasion in hepatocellular carcinoma using contrast-enhanced CT-based deep learning models: a comparative study of transformer, CNN, and MIL
AU - Mianlei Lin
AU - Xiaojun Zheng
AU - Caitao Zhao
AU - Tao Lin
AU - Qishu Zhao
PY - 2026
DA - 2026/11/02
JO - Frontiers in Oncology
DO - 10.3389/fonc.2026.1988882
UR - https://www.frontiersin.org/articles/10.3389/fonc.2026.1988882
AB - ObjectiveMicrovascular invasion (MVI) predicts postoperative recurrence in hepatocellular carcinoma (HCC), yet preoperative prediction remains challenging. This study compared a Transformer, a convolutional neural network (CNN), and a multiple instance learning (MIL) framework for predicting MVI from contrast-enhanced CT.MethodsWe retrospectively enrolled 236 patients with pathologically confirmed HCC. Tumor regions of interest were semi-automatically delineated on arterial-phase CT images, and slices centered on the largest tumor cross-section were sampled to build a 2.5D dataset. Four models—Transformer, MIL, CNN, and a clinical baseline—were trained. Models were compared using the area under the receiver operating characteristic curve (AUC) with the DeLong test; bootstrap resampling assessed the stability of AUC estimates.ResultsThe Transformer achieved the highest AUC of 0.827 (95% CI 0.7585–0.8962), with an accuracy of 0.775, sensitivity of 0.577, and specificity of 0.889, followed by MIL (0.766; 95% CI 0.6875–0.8441), CNN (0.723), and the clinical model (0.715). Bootstrap validation confirmed stable AUC estimates for all models, with the Transformer ranking first (mean 0.827; SD 0.034). The Transformer significantly outperformed the CNN (adjusted P = 0.001) and the clinical model (adjusted P = 0.039); after correction for multiple comparisons, its difference from MIL was of borderline significance (adjusted P = 0.097). No significant differences were observed among the CNN, MIL, and clinical models.ConclusionAmong the four models evaluated, the contrast-enhanced CT-based Transformer predicted MVI most accurately, and its advantage persisted after bootstrap internal validation. These findings concern the ranking of discriminative performance only; the clinical utility of the model is not established, and prospective validation in larger cohorts is required before any clinical application.
SN - 2234-943X
ER -