Atrial fibrillation prediction by combining ECG markers and CMR radiomics

Area of Science:

• Cardiology
• Medical Imaging
• Artificial Intelligence

Background:

• Atrial fibrillation (AF) is the most prevalent cardiac arrhythmia, increasing risks of stroke and death.
• AF is characterized by electro-anatomic changes, typically diagnosed via electrocardiogram (ECG).
• Single-point ECGs may miss paroxysmal AF, necessitating improved diagnostic methods.

Purpose of the Study:

• To develop machine learning models for AF discrimination using ECG and radiomics.
• To characterize AF phenotypes through integrated ECG and imaging data.
• To investigate sex-specific differences in AF-related remodeling.

Main Methods:

• Developed machine learning models integrating ECG features and image-derived radiomics phenotypes.
• Created sex-specific models to explore differential remodeling.
• Compared the performance of the integrated model against ECG alone.

Main Results:

• The combined radiomics-ECG model outperformed ECG alone in AF detection.
• The integrated model showed significantly improved performance in women compared to ECG alone (AUC improved, sensitivity increased).
• ECG performance was lower in women than men, but radiomics addition improved accuracy.

Conclusions:

• Integrative radiomics-ECG models offer enhanced AF detection, particularly beneficial for women.
• Findings provide new insights into sex-based variations in AF-related electro-anatomic remodeling.
• This approach holds potential for earlier and more accurate AF diagnosis.