Machine Learning Identifies Metabolic Dysfunction-Associated Steatotic Liver Disease in Patients With Diabetes...

Area of Science:

• Cardiology
• Hepatology
• Medical Informatics

Background:

• Metabolic dysfunction-associated steatotic liver disease (MASLD) in diabetes mellitus (DM) patients increases cardiovascular disease (CVD) risk.
• MASLD is frequently underdiagnosed in this high-risk population.

Purpose of the Study:

• To develop and validate machine learning (ML) models for assessing MASLD risk in patients with DM.
• To identify key discriminative parameters for MASLD prediction in diabetic patients.

Main Methods:

• Feature selection identified eight discriminative parameters: age, BMI, DM type, liver enzymes, platelet count, hyperuricemia, and metformin use.
• Model performance was evaluated using sensitivity, specificity, ROC curve analysis (AUC), and decision curve analysis (DCA).
• Model generalizability was confirmed using a separate test cohort.

Main Results:

• The ML model achieved high accuracy in identifying MASLD in DM patients (sensitivity 0.75, specificity 0.79; AUC 0.84).
• DCA indicated significant clinical utility, with a net benefit ranging from 30% to 84% threshold probability.
• The model demonstrated generalizability in the test set (sensitivity 0.80, specificity 0.74; AUC 0.81), with unsupervised clustering identifying high-risk subgroups.

Conclusions:

• Machine learning offers a high-performance approach for detecting MASLD in patients with DM.
• This ML tool can improve risk stratification and guide cardiovascular risk prevention strategies for high-risk DM patients with MASLD.