Key result
The lightweight model with fused RNN achieved an overall ECG beat classification accuracy of 99.80% and reduced CPU inference time by 5 times compared to the baseline model.
Why the study?
Accurate analysis of ECG signals is difficult and time-consuming, and existing machine learning and deep learning methods have limitations in model rigidity, complexity, and inference speed.
Does a lightweight deep learning model with fused RNN improve inference speed without loss of accuracy compared to a baseline RNN model for ECG beat classification?
Does a lightweight deep learning model with fused RNN improve inference speed without loss of accuracy compared to a baseline RNN model for ECG beat classification?
Absolute Event Rate: 99.8% vs 99.72%
A lightweight deep learning model with fused RNN achieved cardiologist-level accuracy (99.80%) for ECG beat classification while significantly reducing inference time on CPUs, enabling fast analysis on wearable devices.
May enable on-device ECG analysis; leaves open prospective clinical validation before practice adoption.
BACKGROUND: Electrocardiographic (ECG) monitors have been widely used for diagnosing cardiac arrhythmias for decades. However, accurate analysis of ECG signals is difficult and time-consuming work because large amounts of beats need to be inspected. In order to enhance ECG beat classification, machine learning and deep learning methods have been studied. However, existing studies have limitations in model rigidity, model complexity, and inference speed. OBJECTIVE: To classify ECG beats effectively and efficiently, we propose a baseline model with recurrent neural networks (RNNs). Furthermore, we also propose a lightweight model with fused RNN for speeding up the prediction time on central processing units (CPUs). METHODS: We used 48 ECGs from the MIT-BIH (Massachusetts Institute of Technology-Beth Israel Hospital) Arrhythmia Database, and 76 ECGs were collected with S-Patch devices developed by Samsung SDS. We developed both baseline and lightweight models on the MXNet framework. We trained both models on graphics processing units and measured both models' inference times on CPUs. RESULTS: Our models achieved overall beat classification accuracies of 99.72% for the baseline model with RNN and 99.80% for the lightweight model with fused RNN. Moreover, our lightweight model reduced the inference time on CPUs without any loss of accuracy. The inference time for the lightweight model for 24-hour ECGs was 3 minutes, which is 5 times faster than the baseline model. CONCLUSIONS: Both our baseline and lightweight models achieved cardiologist-level accuracies. Furthermore, our lightweight model is competitive on CPU-based wearable hardware.
No takes yet. Share an insight, caveat, or question.
Jeon et al. (2020) studied Cardiac arrhythmias (n=124). Lightweight deep learning model with fused RNN vs. Baseline model with Vanilla RNN was evaluated on Overall beat classification accuracy. The lightweight model with fused RNN achieved an overall ECG beat classification accuracy of 99.80% and reduced CPU inference time by 5 times compared to the baseline model.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: