Machine-Learning Algorithm for Predicting Fatty Liver Disease in a Taiwanese Population.

Area of Science:

• Hepatology
• Medical Informatics
• Machine Learning

Background:

• Non-alcoholic fatty liver disease (NAFLD) incidence is rising globally, becoming a leading cause of liver disease.
• Early identification of fatty liver disease (FLD) is critical for effective intervention and management.
• Current diagnostic methods may not be sufficient for widespread, early FLD screening.

Purpose of the Study:

• To evaluate the performance of various machine learning algorithms in predicting fatty liver disease (FLD).
• To compare the predictive accuracy of machine learning models against the traditional fatty liver index (FLI).
• To identify key clinical and laboratory factors for FLD prediction using machine learning.

Main Methods:

• A retrospective cross-sectional study involving 31,930 Taiwanese subjects from January 2009 to January 2019.
• Analysis of clinical and laboratory data using five machine learning algorithms: xgBoost, neural network, logistic regression, random forest, and support vector machine.
• Comparison of model performance using metrics such as AUROC, accuracy, F1 score, sensitivity, and specificity, contrasted with the FLI.

Main Results:

• The xgBoost model demonstrated the highest performance with an AUROC of 0.882, accuracy of 0.833, and F1 score of 0.829.
• xgBoost, neural network, and logistic regression models showed significantly higher AUROC than the FLI.
• Body mass index (BMI) was identified as the most crucial feature for predicting FLD.

Conclusions:

• Machine learning algorithms, especially xgBoost, offer superior prediction ability for diagnosing FLD compared to the FLI.
• These algorithms provide substantial advantages for screening individuals at risk of FLD.
• The findings support the integration of machine learning into clinical practice for early FLD detection.