Related Experiment Video
Updated: Jul 11, 2026
12:18
A Machine Learning Approach to Design an Efficient Selective Screening of Mild Cognitive Impairment
Published on: January 11, 2020
Comprehensive and low-cost machine learning models for predicting MASLD prevalence: cross-sectional evidence from the
Azam Doustmohammadian1, Mohammad Farahmand2, Maziar Moradi-Lakeh1
1Gastrointestinal and Liver Diseases Research Center, Iran University of Medical Sciences, Tehran, Iran.
Journal of Health, Population, and Nutrition
|July 7, 2026
Summary
Machine learning models can predict metabolic dysfunction-associated steatotic liver disease (MASLD) using accessible, low-cost indicators. Waist circumference and metabolic factors are key predictors, enabling potential scalable screening for MASLD.
Area of Science:
- Hepatology and Metabolic Disorders
- Artificial Intelligence in Healthcare
- Public Health and Epidemiology
Background:
- Metabolic dysfunction-associated steatotic liver disease (MASLD) is a growing global health concern.
- Early detection of MASLD is challenging, particularly in resource-limited settings.
- Development of accessible screening tools is crucial for managing MASLD prevalence.
Purpose of the Study:
- To develop an interpretable machine learning model for assessing MASLD probability.
- To utilize conventional and non-invasive, low-cost predictors for MASLD detection.
- To explore sex-specific differences in MASLD predictor influence.
Main Methods:
- Employed machine learning algorithms (Random Forest, Gradient Boosting, SVM, k-NN, NN, Recursive Partitioning) on data from 3120 adults.
- Utilized SHapley Additive exPlanations (SHAP) for model interpretability.
- Stratified analyses by sex to identify heterogeneous predictor effects.
Main Results:
- Random Forest and Gradient Boosting models achieved high performance (AUC 0.858 and 0.855).
- Key predictors included composite hepatic indices (Fatty Liver Index, Hepatic Steatosis Index), waist circumference, and adiposity measures.
- Non-invasive models highlighted waist circumference, diastolic blood pressure, age, and lifestyle score as dominant determinants.
Conclusions:
- Visceral adiposity and metabolic dysfunction are strongly associated with MASLD.
- Machine learning models using accessible indicators can effectively identify individuals with MASLD.
- Findings suggest potential for scalable MASLD screening in resource-limited settings.