T2D-LVDD: neural network-based predictive models for left ventricular diastolic dysfunction in type 2 diabetes.

Area of Science:

• Cardiology
• Artificial Intelligence
• Diabetology

Background:

• Cardiovascular disease is a major complication in Type 2 diabetes (T2DM).
• Left ventricular diastolic dysfunction (LVDD) is an early indicator of diabetic cardiac dysfunction.
• Early diagnosis of LVDD is crucial to prevent severe cardiovascular events.

Purpose of the Study:

• To develop a predictive model for LVDD risk in T2DM patients.
• To enable early diagnosis of cardiac dysfunction.
• To identify key risk and protective factors for LVDD.

Main Methods:

• An artificial neural network (ANN) model was trained using patient clinical data.
• The ANN model's performance was compared against logistic regression, random forest, and support vector machine algorithms.
• Interpretability methods were employed to identify LVDD-related features.

Main Results:

• The ANN model demonstrated superior predictive performance compared to classical machine learning methods.
• Key clinical features contributing to LVDD risk were identified.
• A freely accessible web server, LVDD-risk, was developed for risk assessment.

Conclusions:

• The developed ANN model effectively predicts LVDD risk in T2DM patients.
• The LVDD-risk web server provides a valuable tool for early cardiac dysfunction diagnosis.
• Identifying risk factors can guide preventative strategies for diabetic cardiovascular complications.