Key result
A deep neural network for continuous AF monitoring using photoplethysmography achieved an AUC of 0.973, similar to a signal processing model (AUC 0.972), but classified more data (95% vs 62%).
Why the study?
Continuous AF monitoring using wearable PPG is limited by poor signal quality during motion, leading devices to sample infrequently at rest and impeding continuous monitoring or AF quantification.
Does a convolutional deep neural network improve continuous atrial fibrillation monitoring from photoplethysmography compared to a signal processing heuristic in free-living patients?
Observational (n=204)
Does a convolutional deep neural network improve continuous atrial fibrillation monitoring from photoplethysmography compared to a signal processing heuristic in free-living patients?
A deep neural network for continuous PPG monitoring of atrial fibrillation performs similarly to signal processing models in accuracy but classifies significantly more data by better tolerating motion artifacts.
Answers are grounded in this paper and cite their sources.
May increase usable PPG data for continuous AF monitoring without accuracy loss; hypothesis-generating pending prospective validation.
BACKGROUND: Continuous monitoring for atrial fibrillation (AF) using photoplethysmography (PPG) from smartwatches or other wearables is challenging due to periods of poor signal quality during motion or suboptimal wearing. As a result, many consumer wearables sample infrequently and only analyze when the user is at rest, which limits the ability to perform continuous monitoring or to quantify AF. OBJECTIVES: This study aimed to compare 2 methods of continuous monitoring for AF in free-living patients: a well-validated signal processing (SP) heuristic and a convolutional deep neural network (DNN) trained on raw signal. METHODS: We collected 4 weeks of continuous PPG and electrocardiography signals in 204 free-living patients. Both SP and DNN models were developed and validated both on holdout patients and an external validation set. RESULTS: The results show that the SP model demonstrated receiver-operating characteristic area under the curve (AUC) of 0.972 (sensitivity 99.6%, specificity: 94.4%), which was similar to the DNN receiver-operating characteristic AUC of 0.973 (sensitivity 92.2, specificity: 95.5%); however, the DNN classified significantly more data (95% vs 62%), revealing its superior tolerance of tracings prone to motion artifact. Explainability analysis revealed that the DNN automatically suppresses motion artifacts, evaluates irregularity, and learns natural AF interbeat variability. The DNN performed better and analyzed more signal in the external validation cohort using a different population and PPG sensor (AUC, 0.994; 97% analyzed vs AUC, 0.989; 88% analyzed). CONCLUSIONS: DNNs perform at least as well as SP models, classify more data, and thus may be better for continuous PPG monitoring.
No takes yet. Share an insight, caveat, or question.
Antiperovitch et al. (2024) conducted an observational in Atrial fibrillation (n=204). Convolutional deep neural network (DNN) vs. Signal processing (SP) heuristic was evaluated on Receiver-operating characteristic area under the curve (AUC). A deep neural network for continuous AF monitoring using photoplethysmography achieved an AUC of 0.973, similar to a signal processing model (AUC 0.972), but classified more data (95% vs 62%).
Synapse has enriched 10 related papers. Consider them for comparative context: