Artificial Intelligence–Enabled ECG for Diastolic Dysfunction in Congenital Heart Disease: A Novel Risk Stratification Tool

Comprehensive Summary

While clinicians regard invasive haemodynamic evaluations as the most effective tools for the assessment of diastolic function, a lack of access, cost, and procedural risks associated with this strategy must be addressed. Given this, O’Sullivan et. al. aimed to validate an artificial intelligence-enabled electrocardiography (AI-ECG) model for analyzing diastolic dysfunction in patients with adult congenital heart disease (ACHD). The researchers compared their proposed AI-ECG model’s correlation with invasive haemodynamic markers and clinical outcomes. A single-center, retrospective, cohort study was conducted with 6741 patients. The AI-ECG model predicted diastolic dysfunction severity though a scale ranging from 0-3, with increasing severity. Invasive parameters used for comparison to the gold standard included mean right atrial (RA) pressure, pulmonary artery wedge pressure (PAWP), the pulmonary artery, systolic, diastolic, and mean pressures, pulmonary and vascular resistance, and cardiac index. Results indicated the following diastolic function was detected by the model per grade: 65.8%, 4.0%, 19.7%, and 10.5% in grades 0-3, respectively. Invasive haemodynamic statistics followed similar patterns to the model for RA pressure and PAWP. The predicted AI-ECG model provided mortality levels in patients with ACHD and is comparable to invasive measures.

Outcomes and Implications

The AI-ECG model would greatly benefit populations requiring individualized, noninvasive diastolic assessments for ACHD as it provides a method of grading diastolic dysfunction severity. This model would also be clinically appropriate for risk stratification due to its correlation with invasive, gold-standard metrics. Clinical trials are needed to assess the accuracy of prediction in diverse resource communities