Big data reveals insights for lead importance in ECG interpretation

Area of Science:

• Biomedical Engineering
• Cardiology
• Signal Processing

Background:

• Electrocardiogram (ECG) interpretation relies on multiple leads, but their individual importance for specific abnormalities is not always clear.
• Cardiac electrophysiology and expert experience do not always clarify the contribution of each lead.
• Quantifying lead importance is crucial for accurate ECG interpretation and potentially optimizing diagnostic processes.

Purpose of the Study:

• To develop and validate an algorithm for quantifying the importance of individual ECG leads in diagnosing cardiac abnormalities.
• To determine the weight of evidence from each lead during ECG interpretation.
• To identify which leads are most critical for detecting specific ECG abnormalities.

Main Methods:

• An algorithm was created to identify the top K most morphologically similar ECGs in a database for each lead.
• ECG interpretation was performed for each lead using weighted average voting based on similarity.
• The F1 score (sensitivity and positive predictive value) was used to determine the optimal threshold for abnormality detection, defining lead importance.

Main Results:

• The algorithm quantified lead importance across eighteen morphology-based abnormality categories in two databases.
• Results largely confirmed expert knowledge but also revealed novel insights.
• Lead aVR was consistently among the top 6 most important leads for 11-12 abnormality categories, ranking first overall.

Conclusions:

• Lead importance values can guide the selection of critical leads for specific abnormality screening.
• This information may enable the development of simplified ECG monitoring systems, such as those using wearable patches.
• Optimizing lead selection can enhance the efficiency and effectiveness of ECG-based diagnostics.