MRI-based Intra- and Peritumoral Heterogeneity in Hepatocellular Carcinoma for Microvascular Invasion Prediction and...

Area of Science:

• Radiology
• Oncology
• Medical Imaging

Background:

• Hepatocellular carcinoma (HCC) diagnosis and prognosis are challenging.
• Microvascular invasion (MVI) is a key prognostic factor in HCC.
• Quantifying tumor heterogeneity is crucial for improving diagnostic and prognostic models.

Purpose of the Study:

• To develop and evaluate an MRI-based strategy for quantifying intra-tumoral heterogeneity (ITH) and peritumoral heterogeneity (PTH) in HCC.
• To create ITH- and PTH-based models for diagnosing MVI and stratifying prognostic risk in HCC patients.

Main Methods:

• Retrospective analysis of HCC patients (≤5 cm) from three institutions.
• Unsupervised clustering to categorize tumor and peritumoral tissues into habitats on MR images.
• Extraction and quantification of radiomic features to represent ITH and PTH.
• Development of machine learning models (TH_DNN) for MVI diagnosis and survival analysis using Kaplan-Meier curves and Cox regression.

Main Results:

• The TH_DNN model, integrating ITH and PTH features, showed high predictive performance for MVI across datasets (AUC range: 0.82-0.99).
• Patients predicted as MVI-positive by the TH_DNN model had significantly poorer prognosis (overall survival and recurrence).
• Hazard ratios for MVI diagnosis were 2.79 for overall survival and 2.17 for postoperative recurrence.

Conclusions:

• The developed MRI-based strategy effectively quantifies ITH and PTH in HCC.
• This approach enables noninvasive and accurate identification of MVI.
• The strategy is valuable for prognostic risk stratification in HCC patients.