MRI-Based Models Using Habitat Imaging for Predicting Distinct Vascular Patterns in Hepatocellular Carcinoma

Area of Science:

• Radiology and Medical Imaging
• Machine Learning in Oncology
• Hepatocellular Carcinoma (HCC) Research

Background:

• Microvascular invasion (MVI) and vessels encapsulating tumor clusters (VETC) are critical prognostic factors in hepatocellular carcinoma (HCC).
• Accurate preoperative prediction of MVI and VETC remains challenging, impacting treatment decisions and patient outcomes.

Purpose of the Study:

• To develop and validate distinct habitat imaging-based models for predicting MVI and VETC in HCC.
• To integrate these predictive models for enhanced prognosis assessment and patient risk stratification.

Main Methods:

• Multicenter retrospective study involving segmentation of tumor and peritumoral regions from hepatobiliary phase images.
• Feature extraction and analysis using machine learning algorithms to build MVI and VETC predictive models.
• Cox regression analysis to identify independent predictors of early recurrence.

Main Results:

• Developed MVI and VETC prediction models demonstrated high performance in training and external validation cohorts (AUCs ranging from 0.820 to 0.961).
• Patients were successfully stratified into low-, medium-, and high-risk groups based on model predictions.
• Risk group, tumor number, and gender were identified as independent predictors of early recurrence.

Conclusions:

• Habitat imaging-based models enable accurate, noninvasive, preoperative prediction of MVI and VETC in HCC.
• These models provide valuable diagnostic insights and facilitate risk stratification for improved patient management.