Machine-Learning Application for Predicting Metabolic Dysfunction-Associated Steatotic Liver Disease Using Laboratory...

Area of Science:

• Medical Informatics
• Public Health
• Hepatology

Background:

• Metabolic dysfunction-associated steatotic liver disease (MASLD) is a growing global health concern.
• Effective MASLD management relies on early detection and prevention.
• Curative therapies for MASLD are currently limited.

Purpose of the Study:

• To develop and validate machine learning (ML) algorithms for MASLD screening.
• To identify key predictors for MASLD in a large, diverse population.
• To assess the utility of ML in resource-limited settings.

Main Methods:

• Utilized data from the prospective Fasa Cohort Study.
• Employed a two-step predictor selection process combining statistical methods and clinical expertise.
• Compared logistic regression, Naïve Bayes, SVM, and LightGBM algorithms using a 70/30 train/validation split and 5-fold cross-validation.

Main Results:

• Included 6,180 adults (52.7% female); 1364 MASLD cases.
• Logistic regression achieved the highest accuracy (0.88) and AUC (0.92).
• Key predictors identified: waist circumference, BMI, hip circumference, wrist circumference, ALT, cholesterol, glucose, HDL, and blood pressure.

Conclusions:

• ML integration shows promise for MASLD management, especially in rural areas.
• Predictor importance highlights waist circumference and BMI as crucial for MASLD prevention.
• Findings offer valuable insights for MASLD diagnosis, treatment, and prevention strategies.