Machine Learning Algorithms for Predicting Fatty Liver Disease.

Area of Science:

• Medical Informatics
• Machine Learning in Healthcare
• Predictive Diagnostics

Background:

• Fatty liver disease (FLD) is a prevalent condition linked to significant morbidity and mortality.
• Early prediction of FLD is crucial for timely preventive measures, diagnosis, and treatment.
• Developing effective predictive models can aid in managing FLD risk.

Purpose of the Study:

• To develop a machine learning (ML) model for predicting fatty liver disease (FLD).
• To assist healthcare professionals in classifying individuals at high risk of FLD.
• To facilitate novel approaches in FLD diagnosis, management, and prevention.

Main Methods:

• Recruited 3,419 subjects, with 845 screened for FLD.
• Employed various classification models: logistic regression, random forest, ANNs, KNNs, XGBoost, and LDA.
• Assessed predictive accuracy using AUC, sensitivity, specificity, PPV, and NPV.

Main Results:

• Machine learning models demonstrated high predictive accuracy for FLD.
• The XGBoost model achieved the highest accuracy at 0.9415.
• Feature importance analysis identified known FLD risk factors and novel predictors, including hemoglobin.

Conclusions:

• The XGBoost model enables efficient FLD identification in general patients.
• Implementation of this model supports improved prevention strategies.
• Facilitates early treatment and better management of fatty liver disease.