Enhancing tertiary cardiology triage with vectorcardiographic features: a machine learning approach using real-world...

Area of Science:

• Cardiology
• Medical Technology
• Artificial Intelligence in Healthcare

Background:

• Identifying patients needing tertiary cardiac care is crucial for timely intervention.
• Standard ECG features and clinical risk factors have limitations in predicting complex cardiovascular events.
• Global Electrical Heterogeneity (GEH) represents novel electrocardiographic markers.

Purpose of the Study:

• To evaluate if Global Electrical Heterogeneity (GEH) markers enhance the identification of patients requiring tertiary cardiac care.
• To compare the efficacy of GEH alone, combined with machine learning, against standard ECG and clinical factors.
• To assess the performance in a real-world tertiary cardiology population.

Main Methods:

• ECG data and clinical risk factors were collected from patients in a specialized cardiology hospital.
• Global Electrical Heterogeneity (GEH) was derived from ECGs using the Kors Matriz.
• XGBoost decision tree models were trained and optimized using GEH, standard ECG features, and clinical risk factors.

Main Results:

• GEH parameters, including QRST angle and SVG magnitude, were statistically significant (p < 0.001).
• The combined model integrating GEH, standard ECG, and clinical risk factors showed the highest performance (AUC 67.6%).
• Feature importance analysis confirmed the contribution of GEH to the model's predictive power.

Conclusions:

• VCG-derived GEH features show promise in improving the identification of patients needing tertiary cardiac care.
• Integrating GEH into explainable machine learning models enhances predictive capabilities for real-world data.
• Prospective validation is necessary to confirm the clinical utility and integration into care pathways.