Improved ECG Lead Detection and Classification using YOLOv8

The electrocardiogram (ECG) is an essential tool in diagnosing heart conditions, providing critical insights through its 12 individual leads. Accurate analysis of these leads is vital for a precise diagnosis. Traditionally, ECG analysis has predominantly focused on signal-based methodologies. However, integrating image analysis of ECGs alongside signal analysis can significantly enhance diagnostic accuracy and efficiency. To explore this integration, we at Emprotek Software Solutions conducted an experiment utilizing YOLOv8, a stateof-the-art object detection model, trained over 100 epochs to detect ECG leads from ECG report images effectively. Our experimentation utilized the "ECG Images Dataset for Cardiac Patients," a comprehensive collection of ECG images representative of various cardiac conditions. The performance metrics obtained from our model are impressive, with a mean Average Precision (mAP) of 0.992, recall (R) of 1, mAP at 50% IoU (mAP50) of 0.995, and mAP at 50-95% IoU (mAP50-95) of 0.756. These results demonstrate a high degree of accuracy and consistency, aligning well with existing literature in the field of ECG analysis. Furthermore, our study highlights the potential for advanced image analysis techniques to complement traditional signal-based approaches, paving the way for more robust and comprehensive ECG diagnostics. The promising outcomes from our experimentation underscore the importance of continued research and development in this hybrid analysis domain, aiming to enhance the precision and reliability of cardiac health assessments.