Machine Learning and SHAP Value Interpretation for Predicting Hepatic Steatosis Using Vibration-Controlled Transient...

Area of Science:

• Hepatology and Gastroenterology
• Medical Informatics and Machine Learning
• Public Health and Epidemiology

Background:

• Hepatic steatosis (HS), or fatty liver disease, is a prevalent condition marked by fat buildup in liver cells.
• Vibration-controlled transient elastography (VCTE) is used to diagnose HS.
• Machine learning (ML) models offer a promising approach for predicting HS.

Purpose of the Study:

• To develop and validate a machine learning-based predictive model for hepatic steatosis (HS).
• To identify key predictors of HS using advanced interpretation techniques.
• To assess the potential of the developed model for early HS identification and risk stratification.

Main Methods:

• Utilized data from two NHANES cycles (2017-2023) involving 12,177 participants.
• Applied Boruta algorithm and LASSO regression for feature selection, identifying 14 significant predictors.
• Developed and compared six ML models (XGBoost, LR, RF, SVM, MLP, KNN), evaluating performance with AUC and other metrics. SHAP analysis was used for interpretation.

Main Results:

• The XGBoost model demonstrated the highest predictive performance with an AUC of 0.832.
• Key predictors identified by SHAP analysis include Body Mass Index (BMI), glycohemoglobin, age, and HDL cholesterol.
• The model successfully integrated diverse clinical and biochemical data for HS prediction.

Conclusions:

• The XGBoost ML model exhibits strong discriminative ability for predicting hepatic steatosis.
• BMI, glycohemoglobin, age, and HDL cholesterol are identified as the most influential factors in HS prediction.
• This interpretable ML model can potentially enhance early diagnosis and risk stratification of hepatic steatosis.