Compressed Deep Learning to Classify Arrhythmia in an Embedded Wearable Device

Abstract

The importance of an embedded wearable device with automatic detection and alarming cannot be overstated, given that 15–30% of patients with atrial fibrillation are reported to be asymptomatic. These asymptomatic patients do not seek medical care, hence traditional diagnostic tools including Holter are not effective for the further prevention of associated stroke or heart failure. This is likely to be more so in the era of COVID-19, in which patients become more reluctant on hospitalization and checkups. However, little literature is available on this important topic. For this reason, this study developed efficient deep learning with model compression, which is designed to use ECG data and classify arrhythmia in an embedded wearable device. ECG-signal data came from Korea University Anam Hospital in Seoul, Korea, with 28,308 unique patients (15,412 normal and 12,896 arrhythmia). Resnets and Mobilenets with model compression (TensorFlow Lite) were applied and compared for the diagnosis of arrhythmia in an embedded wearable device. The weight size of the compressed model registered a remarkable decrease from 743 MB to 76 KB (1/10000), whereas its performance was almost the same as its original counterpart. Resnet and Mobilenet were similar in terms of accuracy, i.e., Resnet-50 Hz (97.3) vs. Mo-bilenet-50 Hz (97.2), Resnet-100 Hz (98.2) vs. Mobilenet-100 Hz (97.9). Here, 50 Hz/100 Hz denotes the down-sampling rate. However, Resnets took more flash memory and longer inference time than did Mobilenets. In conclusion, Mobilenet would be a more efficient model than Resnet to classify arrhythmia in an embedded wearable device.

Wearable Devices Combined with Artificial Intelligence—A Future Technology for Atrial Fibrillation Detection?

“…Machine learning has many tasks with AF to apply in addition to detection [21], such as risk assessment in real time and AF management. One significant task to tackle with arrhythmias using DNN is detection using embedded or wearable devices [22]. It would have a significant impact on the future if these devices integrated with DNN were part of diagnosing and treating AF.…”

Feelings from the Heart Part II: Simulation and Validation of Static and Dynamic HRV Decrease-Trigger Algorithms to Detect Stress in Firefighters

“…It should be held in mind, however, that firstly, these machine learning approaches do not work online [8]. However, compressed deep learning networks to classify heartbeats and arrythmia were recently developed [38,39]. Hence, compressed deep learning networks to detect stress in everyday life could constitute promising tools in ambulatory research in the future.…”

Electrocardiogram Monitoring Wearable Devices and Artificial-Intelligence-Enabled Diagnostic Capabilities: A Review

“…Due to their ubiquitous availability, most ECG-AI research has been performed using public databases such as the PhysioNet [ 42 ] MIT-BIH Arrhythmia database [ 43 , 44 ] while only a few research groups have independently acquired data from patients. Curated and publicly available datasets include physician annotations that provide a reference for ECG-AI algorithm training ( Table 1 ).…”

“…These approaches utilize fused recurrent neural network (RNN) layers instead of standard RNN layers [ 39 ]. The application of compression [ 44 , 64 ] and conversion techniques (Acc 99.60%) [ 65 ], and model-hardware co-optimization [ 66 ] to reduce the model’s size in terms of computational parameters, resulted in lower memory consumption and inference time. Other techniques to accelerate arrhythmia detection include real-time data compression, signal processing, and data transmission [ 67 , 68 , 69 ].…”