Predictive Efficacy of a Combined Triphasic CT Radiomics and Clinical Feature Model for Ki-67 Expression in...

Area of Science:

• Radiology and Oncological Imaging
• Medical Informatics
• Biomarker Discovery

Background:

• Ki-67 is a crucial biomarker for hepatocellular carcinoma (HCC) aggressiveness and prognosis.
• Accurate preoperative assessment of Ki-67 is essential for risk stratification and personalized treatment planning.
• Current methods for Ki-67 assessment are invasive, highlighting the need for non-invasive alternatives.

Purpose of the Study:

• To develop and validate a prediction model for preoperative Ki-67 expression status in HCC.
• To integrate triphasic contrast-enhanced CT (CECT) radiomics with clinical features.
• To establish a reliable non-invasive tool for HCC risk stratification.

Main Methods:

• Retrospective dual-center study with 213 HCC lesions from 200 patients.
• Dichotomization of Ki-67 expression into high (>20%) and low (≤20%).
• Extraction of radiomic features from arterial, portal venous, and delayed CECT phases; construction of single-phase, multi-phase fusion, clinical, and combined clinical-radiomics fusion models using logistic regression.

Main Results:

• The combined clinical-radiomics fusion model demonstrated robust discrimination with AUCs of 0.866 (training) and 0.824 (internal test).
• External validation showed an AUC of 0.829, significantly outperforming arterial phase (0.713) and radiomics-only (0.722) models.
• The fusion model provided significant incremental value (NRI=0.374, IDI=0.191) and superior clinical net benefit.

Conclusions:

• A fusion model integrating multi-phase CECT radiomics and clinical indicators offers an effective non-invasive method for preoperative Ki-67 prediction in HCC.
• This model can aid in risk stratification and inform individualized treatment strategies.
• The developed tool has the potential to enhance clinical decision-making for HCC management.