Atrial fibrillation detection from raw photoplethysmography waveforms: A deep learning application

Abstract

BACKGROUND Atrial fibrillation (AF), a common cause of stroke, often is asymptomatic. Smartphones and smartwatches can detect AF using heart rate patterns inferred using photoplethysmography (PPG); however, enhanced accuracy is required to reduce false positives in screening populations. OBJECTIVE The purpose of this study was to test the hypothesis that a deep learning algorithm given raw, smartwatch-derived PPG waveforms would discriminate AF from normal sinus rhythm better than algorithms using heart rate alone. METHODS Patients presenting for cardioversion of AF (n 5 51) were given wrist-worn fitness trackers containing PPG sensors (Jawbone Health). Standard 12-lead electrocardiograms over-read by board-certified cardiac electrophysiologists were used as the reference standard. The accuracy of PPG signals to discriminate AF from sinus rhythm was evaluated by conventional measures of heart rate variability, a long short-term memory (LSTM) neural network given heart rate data only, and a deep convolutional-recurrent neural net (DNN) given the raw PPG data. RESULTS From among 51 patients with persistent AF (age 63.6 6 11.3 years; 78% male; 88% white), we randomly assigned 40 to train and 11 to test the algorithms. Whereas logistic regression analysis of heart rate variability yielded an area under the receiver operating characteristic curve (AUC) of 0.717 (sensitivity 0.741; specificity 0.584), the LSTM model given heart rate data exhibited AUC of 0.954 (sensitivity 0.810; specificity 0.921), and the DNN model given raw PPG data yielded the highest AUC of 0.983 (sensitivity 0.985; specificity 0.880). CONCLUSION A deep learning model given the raw PPG-based signal resulted in AF detection with high accuracy, performing better than conventional analyses relying on heart rate series data alone.

Premature Atrial and Ventricular Contraction Detection Using Photoplethysmographic Data from a Smartwatch

“…Indeed, we have shown that detecting PAC/PVC can lead to significant false positive reduction during AF detection [ 40 ]. Other recent published reports [ 14 , 16 , 17 , 18 , 19 , 20 , 21 , 22 , 23 , 24 , 25 , 26 , 27 , 28 , 29 , 30 , 31 , 32 ] have largely focused on AF without accounting for PAC/PVC and consequently their accuracy of AF detection was suboptimal.…”

“…Moreover, the dataset used in this paper did not consider PVC arrhythmia. Many other studies on AF detection using wrist PPG sensors were conducted [17][18][19][20][21][22][23][24][25][26][27][28][29][30][31][32], however, none of them considered PAC or PVC case in their AF detection.…”

Atrial Fibrillation Classification with Smart Wearables Using Short-Term Heart Rate Variability and Deep Convolutional Neural Networks

“…The properties have higher levels of agreement/equivalence when the PPG signals are not excessively situated with motion artifacts. Various predictive and detection models have been implemented using different HRV metrics with standard statistical and machine learning approaches [14][15][16][17][18][19]. However, there are considerably fewer deep learning-based models oriented towards usage in smartphones and wearable devices.…”

Contactless facial video recording with deep learning models for the detection of atrial fibrillation

“…Photoplethysmography (PPG) with pulse waveforms generated from optical sensors of mobile devices has become a new trend and shows sufficient accuracy for the detection of heart rate and other physiological parameters 13 . Recent studies revealed various algorithms with good performance in discriminating AF from sinus rhythm 14 – 17 . Though digital wearables are increasingly popular worldwide, most elderly, who are the main population of AF, are still not used to the application of this high-tech device 18 .…”