Machine Learning in Left Ventricular Hypertrophy Detection: Systematic Review and Meta-Analysis.

Area of Science:

• Cardiology
• Artificial Intelligence
• Medical Informatics

Background:

• Machine learning (ML) is increasingly explored for detecting left ventricular hypertrophy (LVH).
• Existing studies show variable accuracy of ML models for LVH detection, influenced by different variables and algorithms.
• There is a need for systematic evidence on how various ML approaches impact LVH detection accuracy.

Purpose of the Study:

• To systematically assess the diagnostic accuracy of ML approaches for LVH detection.
• To provide evidence for the development of advanced artificial intelligence tools in cardiology.
• To understand the impact of different data types and algorithms on ML model performance for LVH.

Main Methods:

• A systematic literature search was conducted across PubMed, Embase, Cochrane Library, and Web of Science up to November 2025.
• The Prediction Model Risk of Bias Assessment Tool was used for quality assessment.
• Meta-analysis and subgroup analyses were performed on diagnostic 2x2 tables from validation sets, stratified by ML model type and input data (ECG, clinical features, echocardiography).

Main Results:

• Twenty-five studies were analyzed, revealing performance variations in ML models based on input data and algorithms.
• Electrocardiogram (ECG)-based models showed a pooled sensitivity of 0.76 and specificity of 0.84.
• Echocardiography-based models demonstrated a sensitivity range of 0.71-0.94 and specificity of 0.67-0.96; clinical feature models had sensitivity of 0.78 and specificity of 0.71.

Conclusions:

• ML models exhibit moderate accuracy for LVH detection, but evidence is limited and heterogeneity is high.
• Conclusions on ML model accuracy for LVH should be interpreted cautiously due to significant variability.
• Future research should concentrate on developing high-performance ML models utilizing imaging data for more reliable LVH diagnosis.