A Machine Learning Based Framework to Identify and Classify Non-alcoholic Fatty Liver Disease in a Large-Scale...

Area of Science:

• Hepatology
• Medical Informatics
• Public Health

Background:

• Non-alcoholic fatty liver disease (NAFLD) is a prevalent global health concern with limited effective treatments.
• Accurate and efficient screening methods are crucial for managing the growing burden of NAFLD.
• Machine learning (ML) offers potential for developing advanced diagnostic tools in healthcare.

Purpose of the Study:

• To develop and validate machine learning models for the accurate screening of non-alcoholic fatty liver disease (NAFLD).
• To identify key predictive factors for NAFLD using a large-scale dataset.
• To assess the performance of different ML algorithms in NAFLD screening.

Main Methods:

• Utilized data from 304,145 adults undergoing national physical examinations, including questionnaire and physical measurement parameters.
• Applied Absolute Shrinkage and Selection Operator (LASSO) for feature selection from candidate covariates.
• Developed and compared four ML algorithms (XGBoost, etc.) for NAFLD screening, selecting the best-performing classifier.

Main Results:

• XGBoost demonstrated superior performance with an accuracy of 0.880, precision of 0.801, recall of 0.894, F-1 score of 0.882, and AUC of 0.951.
• Key covariates identified for NAFLD screening importance were BMI, age, waist circumference, gender, type 2 diabetes, gallbladder disease, smoking, hypertension, dietary status, physical activity, and dietary preferences.
• The developed ML models facilitate early identification and classification of NAFLD.

Conclusions:

• Machine learning classifiers, particularly XGBoost, provide an effective tool for the early identification and classification of NAFLD.
• The identified covariate importance ranking offers valuable insights for NAFLD prevention and treatment strategies.
• These ML models are particularly beneficial for resource-limited settings to improve NAFLD screening accessibility.