Going beyond Atrial Fibrillation in Arrhythmia Classification from Photoplethysmography Signals

Photoplethysmography (PPG) offers a simple, affordable, and non-invasive method for continuous vascular system monitoring, seamlessly integrated into user-friendly smart devices.Considering the rapid expansion of smart devices, there exists a significant opportunity to extend health monitoring to a broader population.This paper focuses on cardiac arrhythmia (CA) detection from PPG signals.CAs pose significant health risks, often leading to complications such as stroke and heart failure.While most studies focus solely on detecting atrial fibrillation (AF), our research aims to classify six different rhythm types into three classes: Sinus rhythm, AF, and Other.To achieve this goal, we trained a machine learning model -Random Forest.The model was trained on 12 features, which include not only features derived from pulse intervals but also statistical and morphological features.Our results demonstrate an overall accuracy of 0.88 on the test set, indicating the efficacy of our method.Additionally, when tested on a completely independent dataset, the model achieved an accuracy of 0.87.This study highlights the potential of PPG technology to detect various types of CAs beyond AF, offering valuable insights for improved cardiovascular health monitoring.