Can a proposed double branch multimodality-contribution-aware TripNet improve the prediction performance of the microvascular invasion of hepatocellular carcinoma based on small samples?

Preoperative evaluation of microvascular invasion in hepatocellular carcinoma with a radiological feature-based nomogram: a bi-centre study

PURPOSE

To develop a nomogram for preoperative assessment of microvascular invasion (MVI) in hepatocellular carcinoma (HCC) based on the radiological features of enhanced CT and to verify two imaging techniques (CT and MRI) in an external centre.

METHOD

A total of 346 patients were retrospectively included (training, n = 185, CT images; external testing 1, n = 90, CT images; external testing 2, n = 71, MRI images), including 229 MVI-negative patients and 117 MVI-positive patients. The radiological features and clinical information of enhanced CT images were analysed, and the independent variables associated with MVI in HCC were determined by logistic regression analysis. Then, a nomogram prediction model was constructed. External validation was performed on CT (n = 90) and MRI (n = 71) images from another centre.

RESULTS

Among the 23 radiological and clinical features, size, arterial peritumoral enhancement (APE), tumour margin and alpha-fetoprotein (AFP) were independent influencing factors for MVI in HCC. The nomogram integrating these risk factors had a good predictive effect, with AUC, specificity and sensitivity values of 0.834 (95% CI: 0.774-0.895), 75.0% and 83.5%, respectively. The AUC values of external verification based on CT and MRI image data were 0.794 (95% CI: 0.700-0.888) and 0.883 (95% CI: 0.807-0.959), respectively. No statistical difference in AUC values among training set and testing sets was found.

CONCLUSION

The proposed nomogram prediction model for MVI in HCC has high accuracy, can be used with different imaging techniques, and has good clinical applicability.

MVI-TR: A Transformer-Based Deep Learning Model with Contrast-Enhanced CT for Preoperative Prediction of Microvascular Invasion in Hepatocellular Carcinoma

In this study, we considered preoperative prediction of microvascular invasion (MVI) status with deep learning (DL) models for patients with early-stage hepatocellular carcinoma (HCC) (tumor size ≤ 5 cm). Two types of DL models based only on venous phase (VP) of contrast-enhanced computed tomography (CECT) were constructed and validated. From our hospital (First Affiliated Hospital of Zhejiang University, Zhejiang, P.R. China), 559 patients, who had histopathological confirmed MVI status, participated in this study. All preoperative CECT were collected, and the patients were randomly divided into training and validation cohorts at a ratio of 4:1. We proposed a novel transformer-based end-to-end DL model, named MVI-TR, which is a supervised learning method. MVI-TR can capture features automatically from radiomics and perform MVI preoperative assessments. In addition, a popular self-supervised learning method, the contrastive learning model, and the widely used residual networks (ResNets family) were constructed for fair comparisons. With an accuracy of 99.1%, a precision of 99.3%, an area under the curve (AUC) of 0.98, a recalling rate of 98.8%, and an F1-score of 99.1% in the training cohort, MVI-TR achieved superior outcomes. Additionally, the validation cohort's MVI status prediction had the best accuracy (97.2%), precision (97.3%), AUC (0.935), recalling rate (93.1%), and F1-score (95.2%). MVI-TR outperformed other models for predicting MVI status, and showed great preoperative predictive value for early-stage HCC patients.