A Novel Smartphone-based & Personalized Atrial Fibrillation Detection: A Preliminary Study

Abstract

This paper aims to propose a novel system that can be used for personalized atrial fibrillation (AF) detection using smartphone photoplethysmogram (PPG) signals. First, we detect atrial fibrillation (AF) signals from normal heart rhythm signals, and then the AF smartphone PPG signals are used for personalized AF detection. We extracted 19 features from the fiducial and non-fiducial information of smartphone PPG signals. These features were used for both classifying AF signals from normal signals and personalized AF detection of each subject. The ensemble algorithms with the boosting and bagging functions were used for both the AF detection from normal and personalized AF detection processes. We achieved 100% accuracy for detecting AF signals from normal signals and 96.08%. for personalized AF detection. These preliminary results indicate that our proposed system can be used for personalized AF detection and management which has been recently gained attention from researchers.