Prediction of Paroxysmal Atrial Fibrillation With Incorporating Genomic Information Into AI-Based ECG Analysis

Area of Science:

• Cardiology
• Artificial Intelligence in Medicine
• Genetics and Genomics

Background:

• Predicting paroxysmal atrial fibrillation (PAF) using artificial intelligence (AI) models from sinus rhythm electrocardiograms (ECGs) is an evolving field.
• The added value of integrating genetic information, such as polygenic risk scores (PRS), and blood biomarkers into AI-ECG models for PAF prediction remains unclear.

Purpose of the Study:

• To investigate the potential utility of polygenic risk scores (PRS) and specific blood biomarkers in enhancing AI-based ECG (AI-ECG) analysis for predicting paroxysmal atrial fibrillation (PAF).
• To evaluate the combined predictive performance of AI-ECG scores and PRS for PAF.

Main Methods:

• A cohort of 2,128 patients from seven Japanese institutions was analyzed.
• AI-ECG scores and polygenic risk scores (PRS) were calculated using established algorithms.
• Levels of high-sensitivity C-reactive protein, N-terminal pro-B-type natriuretic peptide, and cell-free DNA (cfDNA) were measured.

Main Results:

• AI-ECG scores demonstrated strong predictive performance for PAF (AUC-ROC: 0.882).
• Polygenic risk scores (PRS) also showed significant association with PAF (AUC-ROC: 0.655).
• While combining AI-ECG and PRS did not improve the area under the receiver-operating characteristic curve (AUC-ROC), it significantly enhanced risk reclassification (Net Reclassification Improvement: 0.467) and integrated discrimination (Integrated Discrimination Improvement: 0.028). Cell-free DNA showed a limited contribution.

Conclusions:

• Genetic information, specifically PRS, may offer complementary data to improve risk stratification for paroxysmal atrial fibrillation (PAF) prediction when used alongside AI-ECG.
• The integration of genetic insights holds promise for refining AI-driven cardiovascular risk assessment tools.