Novel Robust-to-Motion-Artifact Detection of Atrial Fibrillation Based on PPG Only

Atrial fibrillation (AFib) is a prevalent heart rhythm disorder linked to severe health risks such as stroke and heart failure. Photoplethysmography (PPG) has emerged as a promising method for AFib detection, owing to its non-invasiveness and wide-ranging applicability. However, the precise analysis of PPG signals remains challenging due to the introduction of motion artifacts, which improve the quality and reliability of the measurements. In response to this challenge, this study proposes a novel approach for robust AFib detection using a 1-dimensional convolutional neural network (1D-CNN) model, explicitly designed to mitigate the effects of motion artifacts without necessitating additional sensors. The proposed method's central innovation lies in its focus on the analysis of single-cycle waveforms. As a critical step in the process, a quality check model was implemented to scrutinize the PPG signal quality. This quality check model achieved an impressive accuracy of 98.26%, sensitivity of 99.09%, and specificity of 97.50%. By systematically removing the cycles that failed to meet the quality criteria set by the model, the accuracy of AFib detection was significantly enhanced, leading to a remarkable increase in detection accuracy from 85.5% to 97.50^%