Triphasic CT Radiomics Model for Preoperative Prediction of Hepatocellular Carcinoma Pathological Grading

Area of Science:

• Radiology
• Oncology
• Medical Imaging

Background:

• Hepatocellular carcinoma (HCC) diagnosis and prognosis rely on accurate pathological markers.
• Noninvasive methods for predicting these markers are crucial for clinical decision-making.

Purpose of the Study:

• To develop and validate a triphasic CT-based radiomics model for predicting multiple critical pathological markers in HCC.
• To assess the model's performance in grading Edmondson-Steiner (Ed), Microvascular invasion (MVI), and Satellite nodule (SN).

Main Methods:

• A retrospective study of 174 HCC patients (187 lesions) using 2264 radiomic features from arterial, venous, and delayed phase CT images.
• Feature selection using mRMR, SelectKBest, and LASSO algorithms.
• Development of single-phase and triphasic fusion models using logistic regression and SVM classifiers.

Main Results:

• The triphasic fusion model demonstrated superior performance in predicting Ed, MVI, and SN grading.
• Achieved high AUCs in both testing (0.890-0.829) and validation (0.836-0.810) datasets.
• Fusion model outperformed individual single-phase models.

Conclusions:

• The triphasic CT radiomics model offers a noninvasive tool for preoperative prediction of HCC pathological grading.
• Enhances diagnostic accuracy for clinical decision-making and prognostic evaluation.
• Facilitates improved patient management strategies for HCC.