Research on a Lightweight Arrhythmia Classification Model Based on Knowledge Distillation for Wearable Single-Lead ECG Monitoring Systems

Abstract

Arrhythmias are among the diseases with high mortality rates worldwide, causing millions of deaths each year. This underscores the importance of real-time electrocardiogram (ECG) monitoring for timely heart disease diagnosis and intervention. Deep learning models, trained on ECG signals across twelve or more leads, are the predominant approach for automated arrhythmia detection in the AI-assisted medical field. While these multi-lead ECG-based models perform well in automatic arrhythmia detection, their complexity often restricts their use on resource-constrained devices. In this paper, we propose an efficient, lightweight arrhythmia classification model using a knowledge distillation technique to train a student model from a teacher model, tailored for embedded intelligence in wearable devices. The results show that the student model achieves 96.32% accuracy, which is comparable to the teacher model, with a remarkable compression ratio that is 1242.58 times smaller, outperforming other lightweight models. Enabled by the proposed model, we developed a wearable ECG monitoring system based on the STM32F429 Discovery kit and ADS1292R chip, achieving real-time arrhythmia detection on small wearable devices.

Keywords: arrhythmia classification; edge intelligence; electrocardiogram (ECG); embedded system; knowledge distillation (KD); microcontroller.

Figure 1

Model architecture and knowledge distillation process. It illustrates the knowledge distillation process, wherein the information acquired by the teacher model is efficiently conveyed to the student model, allowing it to attain superior performance with a more streamlined design.

Figure 2

Confusion matrices for the teacher model under different configurations. (a) Confusion matrix for the complete teacher model; (b) confusion matrix for the teacher model without ResNet blocks; (c) confusion matrix for the teacher model without LSTM layers; (d) confusion matrix for the teacher model without SENet blocks.

Figure 3

Training and validation curves for the student model. (a) Accuracy curve of the student model during training and validation; (b) loss curve of the student model during training and validation.

Figure 4

Confusion matrices for the student model without and with knowledge distillation. (a) Confusion matrix for the student model without knowledge distillation; (b) confusion matrix for the student model with knowledge distillation.

Figure 5

Wearable ECG monitoring system block diagram.

Figure 6

Process of integrating the student model into the STM32F4 board with STM32Cube.AI.

Figure 7

The real-time ECG monitoring and arrhythmia detection system.