Prediction and Staging of Hepatic Fibrosis in Children with Hepatitis C Virus: A Machine Learning Approach

Area of Science:

• Pediatric Hepatology
• Medical Informatics
• Machine Learning in Medicine

Background:

• Chronic hepatitis C (CHC) in children requires accurate fibrosis staging for effective management.
• Non-invasive markers like APRI (aspartate aminotransferase-to-platelet ratio) and FIB-4 (fibrosis score) are crucial for assessing liver fibrosis.
• Existing non-invasive methods may require optimization for pediatric populations.

Purpose of the Study:

• To develop an intelligent diagnostic system using machine learning for data cleansing.
• To build an intelligent model for predicting and staging liver fibrosis in pediatric CHC.
• To establish new cutoff values for APRI and FIB-4 scores in this cohort.

Main Methods:

• Utilized Random Forest (RF) for data cleansing and feature selection.
• Applied RF, APRI, and FIB-4 for fibrosis prediction and staging on a cleaned dataset.
• Evaluated model performance using Area Under the ROC Curve (AUC) on 166 Egyptian children with CHC.

Main Results:

• Random Forest (RF) demonstrated high predictive accuracy with AUCs ranging from 0.822 to 0.903.
• APRI and FIB-4 scores also achieved significant AUCs, indicating good performance in fibrosis staging.
• The study identified specific AUC values for predicting any fibrosis, advanced fibrosis, and differentiating fibrosis stages.

Conclusions:

• Machine learning, particularly RF, is a valuable tool for non-invasive liver fibrosis assessment in pediatric CHC.
• The newly derived cutoff values for APRI and FIB-4 showed good performance and consistency with prior research.
• The findings support the integration of machine learning and optimized non-invasive markers for improved fibrosis diagnosis in children.