End-to-end Deep Learning from Raw Sensor Data: Atrial Fibrillation\n Detection using Wearables

2018/07/27 by Igor Gotlibovych, Stuart L. Crawford, Gotlibovych, Igor +15 · 1 citation
Engineering · Medicine · Neuroscience · #Non-Invasive Vital Sign Monitoring #ECG Monitoring and Analysis #EEG and Brain-Computer Interfaces

paper · pdf · doi:10.48550/arxiv.1807.10707

Abstract

We present a convolutional-recurrent neural network architecture with long\nshort-term memory for real-time processing and classification of digital sensor\ndata. The network implicitly performs typical signal processing tasks such as\nfiltering and peak detection, and learns time-resolved embeddings of the input\nsignal. We use a prototype multi-sensor wearable device to collect over 180h of\nphotoplethysmography (PPG) data sampled at 20Hz, of which 36h are during atrial\nfibrillation (AFib). We use end-to-end learning to achieve state-of-the-art\nresults in detecting AFib from raw PPG data. For classification labels output\nevery 0.8s, we demonstrate an area under ROC curve of 0.9999, with false\npositive and false negative rates both below 2\× 10^-3. This\nconstitutes a significant improvement on previous results utilising\ndomain-specific feature engineering, such as heart rate extraction, and brings\nlarge-scale atrial fibrillation screenings within imminent reach.\n

Citations

• Modern hierarchical, agglomerative clustering algorithms
• Gradient-based learning applied to document recognition
• A guide to convolution arithmetic for deep learning
• Cardiologist-Level Arrhythmia Detection with Convolutional Neural Networks
• Time Series Classification from Scratch with Deep Neural Networks: A Strong Baseline

Cited by

• A systematic evaluation of uncertainty quantification techniques in deep learning: a case study in photoplethysmography signal analysis

Related

• DeepBeat: A multi-task deep learning approach to assess signal quality and arrhythmia detection in wearable devices
• SiamAF: Learning Shared Information from ECG and PPG Signals for Robust Atrial Fibrillation Detection
• Ambulatory Atrial Fibrillation Monitoring Using Wearable Photoplethysmography with Deep Learning
• SQUWA: Signal Quality Aware DNN Architecture for Enhanced Accuracy in Atrial Fibrillation Detection from Noisy PPG Signals
• Detection of Paroxysmal Atrial Fibrillation using Attention-based Bidirectional Recurrent Neural Networks