Avaliação do desempenho de PPG versus ECG na identificação de fibrilação atrial utilizando intervalos RR

Abstract

Atrial fibrillation (AF) is the most common supraventricular arrhythmia, characterized by an irregular heart rhythm due to electrical disorganization of the atria, associated with high morbidity and mortality, especially in the elderly. Although traditional diagnosis is made through electrocardiogram (ECG), wearable devices that use photoplethysmography (PPG) may enable continuous and accessible heart rate monitoring. Thus, this study aims to investigate the feasibility of PPG in detecting AF by comparing heart rate variability metrics between ECG and PPG to identify the most accurate ones. Simultaneous ECG and PPG signals from 32 patients were analyzed, 16 during AF episodes and 16 in sinus rhythm, previously processed using the Python library NeuroKit2, including filtering, peak detection, and quality assessment. Five RR variability metrics (pRR32, pRR40, pRR48, pRR72, and pRR3.25%) were evaluated for sensitivity, specificity, ROC curves, and AUC values, with thresholds defined by the Youden index. The results showed that ECG presented greater accuracy and consistency in peak and AF detection, while PPG maintained good sensitivity but with a higher number of false positives due to signal variability. The pRR3.25% metric stood out for balancing sensitivity and specificity in both modalities. However, more sensitive metrics such as pRR40 or pRR72 may be advantageous for preliminary screening with PPG, prioritizing the detection of true positives even with a slight increase in false positives. Although less accurate, PPG demonstrated practicality and clinical applicability, highlighting the need for improvement in quality assessment criteria and enhancement of algorithms, as well as analyses on a larger database to increase its reliability.