Machine learning models for predicting metabolic dysfunction-associated steatotic liver disease prevalence using...

Area of Science:

• Hepatology
• Medical Informatics
• Machine Learning

Background:

• Metabolic dysfunction-associated steatotic liver disease (MASLD) poses a significant global health challenge.
• Current MASLD diagnostic methods are complex and require specialized facilities.
• Early screening and intervention are crucial for improving MASLD patient prognosis.

Purpose of the Study:

• To assess the feasibility of using machine learning (ML) models for large-scale MASLD screening.
• To develop predictive models for MASLD based on demographic and clinical data.
• To evaluate the accuracy and generalizability of ML models for MASLD detection.

Main Methods:

• 10,007 outpatients underwent transient elastography to form a derivation cohort.
• Eight demographic and clinical factors were used to build ML models for MASLD prediction.
• Ten ML algorithms were employed and externally validated on the NHANES 2017-2023 datasets.

Main Results:

• Machine learning models demonstrated robust predictive capabilities in both hospital and external cohorts.
• Logistic Regression (LR) achieved high accuracy (0.711-0.728) and Area Under the Curve (AUC) (0.798-0.806).
• External validation showed consistent performance, with LR, MLP, and XGBoost yielding AUCs of 0.831, 0.823, and 0.784, respectively.

Conclusions:

• ML models utilizing demographic and clinical data can accurately screen for MASLD in the general population.
• This ML-driven approach enhances the feasibility, accessibility, and compliance of MASLD screening.
• The study provides an effective tool for large-scale health assessments and early MASLD intervention strategies.