Related Experiment Videos

Cross-Model Explainability Consistency in Hepatitis C Stage Classification: A SHAP, LIME, and Counterfactual Analysis

Khalid Alalawi1

  • 1College of Computer Science and Engineering (CCSE), Taibah University, Medina 42353, Saudi Arabia.

Insights

Machine learning models can stage Hepatitis C virus (HCV) progression. The study found Cholinesterase (CHE) is a key marker for Cirrhosis, consistent across multiple models and explanation techniques.

Area of Science:

  • Hepatology and Machine Learning
  • Biochemical marker identification for liver disease staging

Background:

  • Hepatitis C virus (HCV) affects 50 million globally, progressing through silent hepatic stages.
  • Existing machine learning staging methods lack clinician-facing explanations and cross-architecture validation.
  • Need for interpretable AI in understanding complex disease progression.

Purpose of the Study:

  • To develop and evaluate machine learning models for staging Hepatitis C virus (HCV) infection.
  • To identify and validate stage-specific biochemical markers using explainable AI techniques.
  • To assess the consistency of feature importance across different model architectures.

Main Methods:

  • Trained five models (Logistic Regression, Random Forest, XGBoost, LightGBM, SVM) on the UCI HCV dataset.
  • Applied explainable AI methods: SHAP, LIME, and DiCE counterfactuals for feature interpretation.
  • Utilized leakage-controlled pipelines, five-fold cross-validation, and cross-model Spearman agreement analysis.

Main Results:

  • Boosted-tree models (XGBoost, LightGBM) showed robust performance in cross-validation (macro-F1 ~0.65).
  • SHAP consistently identified AST, GGT, CHE, and ALP as important features across tree-based models.
  • Cholinesterase (CHE) emerged as a leading Cirrhosis marker, with high consistency between SHAP and LIME explanations.

Conclusions:

  • The developed framework successfully identifies stage-specific biochemical patterns in HCV progression.
  • Convergent evidence supports Cholinesterase (CHE) as a potential marker for advanced HCV-related Cirrhosis.
  • Cross-model validation of AI explanations enhances confidence in identified disease markers.

Related Concept Videos