Mobile Application for Heart Diseases Diagnosis based on the PPG Signals and Deep Learning

Atrial fibrillation (AF) is a cardiac arrhythmia that affects numerous individuals annually. This condition raises the risk of strokes, heart failure, and even mortality. While medical-grade electrocardiogram (ECG) devices provide precise analysis, they are expensive and require clinical settings. Recent advancements in general-purpose smartphones and wearable technology with photoplethysmography (PPG) sensors have improved accessibility for diagnosing AF for most people. This study aims to develop a single model capable of accurately classifying AF using PPG signals obtained through a mobile phone camera. A deep convolutional neural network is proposed to classify 15-second heart rhythms as either normal sinus rhythm or atrial fibrillation based on captured images from the camera. The model is trained using a dataset and evaluated on a separate dataset containing unseen PPG signals from different individuals. The achieved performance measures indicate an accuracy of 99.01%. These results demonstrate significant potential for easily classifying Atrial fibrillation disease through mobile phones.