Automated detection of atrial fibrillation using R-R intervals and multivariate-based classification

• Published in: Journal of electrocardiology Vol. 49; no. 6; pp. 871 - 876
• Main Authors: Kennedy, Alan, Finlay, Dewar D., Guldenring, Daniel, Bond, Raymond R., Moran, Kieran, McLaughlin, James
• Format: Journal Article
• Language: English
• Published: United States Elsevier Inc 01.11.2016; Elsevier Science Ltd
• Subjects: Adolescent; Adult; Aged; Aged, 80 and over; Algorithms; Atrial fibrillation; Atrial Fibrillation - diagnosis; Automation; Cardiac arrhythmia; Cardiovascular; Diagnosis, Computer-Assisted - methods; Electrocardiography; Electrocardiography - methods; Female; Heart Rate Determination - methods; Humans; Male; Medical screening; Middle Aged; Multivariate analysis; Pattern Recognition, Automated - methods; R-R intervals; Reproducibility of Results; Sensitivity and Specificity; Young Adult; Atrial fibrillation; Algorithms; R-R intervals; Atrial Fibrillation
• ISSN: 0022-0736; 1532-8430
• DOI: 10.1016/j.jelectrocard.2016.07.033

Automated detection of AF from the electrocardiogram (ECG) still remains a challenge. In this study, we investigated two multivariate-based classification techniques, Random Forests (RF) and k-nearest neighbor (k-nn), for improved automated detection of AF from the ECG. We have compiled a new database from ECG data taken from existing sources. R-R intervals were then analyzed using four previously described R-R irregularity measurements: (1) the coefficient of sample entropy (CoSEn), (2) the coefficient of variance (CV), (3) root mean square of the successive differences (RMSSD), and (4) median absolute deviation (MAD). Using outputs from all four R-R irregularity measurements, RF and k-nn models were trained. RF classification improved AF detection over CoSEn with overall specificity of 80.1% vs. 98.3% and positive predictive value of 51.8% vs. 92.1% with a reduction in sensitivity, 97.6% vs. 92.8%. k-nn also improved specificity and PPV over CoSEn; however, the sensitivity of this approach was considerably reduced (68.0%).