Transfer Learning for the Prediction of Atrial Fibrillation and Sleep Apnea
Inexpensive and highly portable medical sensing technology has the potential to deliver savings for healthcare providers, through screening and early detection. However, the data necessary to support machine learning models on these newly developed sensor modalities - sensor streams together with time-aligned, expert-annotated labels - is often difficult and expensive to obtain. By contrast, high-quality data suitable for machine learning on better-established modalities, such as electrocardiogram (ECG), is abundant and often free. We propose a method allowing for the use of such easy-to-come-by data in building models on photoplethysmography (PPG) sensor modalities. With this method, data requirements are reduced to streams of PPG data time-aligned with readings from ECG together with labelled outcome for ECG. We demonstrate the method by building models for atrial fibrillation and sleep apnea based on data from a wrist-worn PPG sensor, with the only labels coming from publicly available ECG data. We find that the models developed using the transfer learning approach outperform models trained directly on the PPG sensor and are competitive with state-of-the-art ECG-based models.
3 Key Takeaways:
*Deep learning outperforms traditional machine learning on heart rate variability features in predicting the presence/absence of atrial fibrillation in real-time.
*For sleep apnea, the pre-trained model’s predictions provide a substantially better learning signal than the clinician-provided labels, and that this teacher-student technique significantly outperforms both a naive application of supervised deep learning and a label-supervised version of domain adaptation.
*These applications demonstrate that our wrist-worn device can provide close to clinical-grade accuracy for the real-time prediction of atrifibrlation and sleep apnea.
Dr Jessie Li received her DPhil in Computational Genomics from the University of Oxford. After that she worked at a university spin-off as a statistical geneticist and two healthcare technology startups as a data scientist, most recently serving as Head of Data Science at all.health. She contributed to the understanding of the genetic cause of major depressive disorder during her academic career which could potentially shed light on new diagnostic and treatment solutions. At all.health, she has led the development of a number of machine learning models for disease detection using photoplethysmography data from the company’s proprietary wrist-worn device. This provides close to clinical grade accuracy on a number of conditions using continuously monitored patient data.