A novel ECG detector performance metric and its relationship with missing and false heart rate limit alarms

Area of Science:

• Biomedical Engineering
• Cardiovascular Technology
• Signal Processing

Background:

• Traditional electrocardiogram (ECG) beat detector performance is assessed over long intervals (30 minutes).
• However, incorrect detections within short intervals (10 seconds) are critical for triggering false heart rate limit alarms (tachycardia/bradycardia).

Purpose of the Study:

• To propose and evaluate a novel performance metric for ECG beat detectors based on the distribution of incorrect detections within a short interval.
• To assess the relationship between this new metric and the rate of incorrect heart rate limit alarms.

Main Methods:

• Six ECG beat detectors were evaluated using both traditional long-interval (30-minute) sensitivity and positive predictive value metrics.
• A new short-interval metric, Area Under the empirical cumulative distribution function (AUecdf) for 10-second sensitivity and positive predictive value, was calculated.
• False heart rate limit and asystole alarm rates were determined using a separate ECG database.
• Spearman's rank correlation was used to analyze the relationship between performance metrics and alarm rates.

Main Results:

• Both long-interval sensitivity (30min) and short-interval AUecdf for sensitivity showed significant correlation with false alarm rates (ρ=-0.8 and ρ=0.9, respectively; p<0.05).
• While 30-minute sensitivity grouped detectors with the lowest false alarm rates, AUecdf for sensitivity provided additional discriminatory power to identify detectors with the highest false alarm rates.

Conclusions:

• Short-interval performance metrics for ECG beat detectors offer valuable insights into their propensity to generate incorrect heart rate limit alarms.
• These novel metrics may improve the assessment and selection of ECG beat detection algorithms for clinical applications.