Artificial Intelligence in the Early Detection of Structural Heart Disease via Wearable ECGs

Abstract

Background:
Structural heart diseases (SHD) are hypertrophic heart, valvular, and left-ventricular heart inabilities that cannot be effectively identified
until later phases. Long-term and real-life cardiac monitoring is now accessible in form of wearable ECG devices capable of collecting
and display the data in large amounts, but it is hard to comprehend. The remedy is the artificial intelligence (AI), which can recognize
small electrical signs of SHD that can be difficult to observe during the review of the usual ECG.
Objective:
To determine whether AI-based analysis of wearable ECG among the vulnerable population to the early detection of structural heart
disease is effective.
Method:
Compared to the single-lead and multi-lead ECGs of wearable devices, the simulation of machine-learning models was done using a
literature-based framework and pilot data. Transformer architecture, neural networks and feature-extraction algorithms had been trained
on features that are correlated with echocardiography abnormalities. A comparison of model performance in terms of clinician
interpretation and the standard ECG criteria was done.
Results:
The AI based wearable ECG analysis increased the sensitivity in the detection of SHD related electrical activities like the weak
repolarization changes, micro voltages detection, and conduction delays. It was shown that models were more efficient than the manual
interpretation and identified the high-risk patients earlier than the conventional diagnostic pathways.
Conclusion:
AI implementation in wearable ECGs enhances the procedure of identifying structural heart disease at initial stages that is a large-scale
solution to risk stratification and prompt action. The practice can play a key role in enhancing the clinical outcomes in better diagnosis
and follow up.