Hybrid Bi-LSTM-Spiking Neural Network for Energy-Efficient ECG Arrhythmia Classification

Q: What is the clinical evidence from this study?

Population: ECG Arrhythmia (n=200). Intervention: Hybrid Bi-LSTM-Spiking Neural Network vs. Standalone Spiking Neural Network. Primary outcome: Classification test accuracy.

Q: What does this research mean for the field?

A hybrid model combining a Bi-LSTM layer with leaky integrate-and-fire spiking neural network layers achieves 96.88% test accuracy in ECG arrhythmia classification, outperforming standalone spiking neural networks by 5.21 percentage points while enabling event-driven processing. Novelty: ClaimNovelty.METHODOLOGICAL. Consensus alignment: ConsensusAlignment.NEUTRAL.

Nor Amalia Dayana Mohamad Noor; Wong Yan Chiew; Zarina Mohd Noh; Ranjit Singh Sarban Singh; Cheng Xiang

doi:10.54554/jtec.2026.18.03.002

September 30, 2026Journal of Telecommunication Electronic and Computer Engineering (JTEC)Open Access

Hybrid Bi-LSTM-Spiking Neural Network for Energy-Efficient ECG Arrhythmia Classification

View Full Paper

Ask AI

Bookmark

Share

Key result

Hybrid Bi-LSTM-SNN improves ECG arrhythmia classification accuracy by ~5 percentage points over standalone SNN.

n=200

Why the study?

Wearable ECG monitoring requires a classifier that can operate within a limited power budget while maintaining high classification performance.

Key Points

To develop an energy-efficient ECG arrhythmia classification architecture suitable for wearable devices by combining bidirectional recurrent memory with event-driven spiking neurons.
Coupled a 64-unit bidirectional long short-term memory (Bi-LSTM) layer with two leaky integrate-and-fire (LIF) spiking layers trained using a fast-sigmoid surrogate gradient and backpropagation through time.
Evaluated 20 hyperparameter configurations across 60-40, 70-30, and 80-20 training-testing partitions.
Configuration 3 on an 80-20 split (membrane decay parameter 0.85, learning rate 0.00020) achieved 97.64% training accuracy and 96.88% test accuracy, surpassing the standalone SNN baseline of 91.67% by 5.21 percentage points.
Achieved a precision of 0.93 and recall of 1.00 for Class 0 (normal), alongside a precision of 1.00 and recall of 0.95 for Class 1 (abnormal).

Structured PICO

P

Population

200 digitized ECG records (80 normal, 120 abnormal) from the PTB database evaluated to test a machine learning classifier.

I

Intervention

Hybrid bidirectional long short-term memory (Bi-LSTM) network combined with two leaky integrate-and-fire (LIF) spiking neural network layers.

C

Comparator

Standalone Spiking Neural Network (SNN) and existing Convolutional Neural Network (CNN) architectures.

O

Outcome

ECG classification accuracy on a held-out test partition.

Main Result

Absolute Event Rate: 96.88% vs 91.67%

A hybrid Bi-LSTM and spiking neural network model achieved 96.88% accuracy in ECG arrhythmia classification, providing a high-performance and potentially energy-efficient architecture for wearable cardiac monitoring.

Limitations

Small sample size
Actual energy savings cannot be established until evaluated on intended hardware
Requires validation in a larger, patient-independent cohort
Small sample size of ECG records
Actual energy savings cannot be established until the model is evaluated on hardware.

Implication

May support energy-efficient wearable ECG monitoring; leaves open prospective clinical validation before practice adoption.

Abstract

Wearable ECG monitoring requires a classifier that can operate within a limited power budget. This study therefore combined a bidirectional long short-term memory (Bi-LSTM) layer, which processed the waveform in both temporal directions, with two leaky integrate-and-fire (LIF) spiking layers. The recurrent output was converted into sparse events, while a fast-sigmoid surrogate gradient enabled training through the nondifferentiable spike function using backpropagation through time. Twenty hyperparameter configurations were evaluated using 60-40, 70-30, and 80-20 training-testing partitions. The best configuration, Configuration 3 using the 80-20 partition, comprised one 64-unit Bi-LSTM layer, a membrane decay parameter of 0.85, and a learning rate of 0.00020. It achieved a training accuracy of 97.64% and a test accuracy 96.88%, exceeding the reporting accuracy of the standalone SNN by 5.21 percentage points. Thus, the model maintained high ECG classification performance while shifting its later-stage processing to event-driven spike activity.

Discussion

No takes yet. Share an insight, caveat, or question.

Cite This Study

Noor et al. (2026) studied ECG Arrhythmia (n=200). Hybrid Bi-LSTM-Spiking Neural Network vs. Standalone Spiking Neural Network was evaluated on Classification test accuracy. The hybrid Bi-LSTM-SNN achieved a test accuracy of 96.88% for ECG arrhythmia classification, outperforming a standalone SNN baseline of 91.67%.

synapsesocial.com/papers/6ace6db8b52af824ab02bb22 https://doi.org/10.54554/jtec.2026.18.03.002

View Full Paper

Ask AI

Bookmark

Share

Hybrid Bi-LSTM-Spiking Neural Network for Energy-Efficient ECG Arrhythmia Classification

Key Points

Structured PICO

Main Result

Limitations

Implication

Abstract

Discussion

Cite This Study

Hybrid Bi-LSTM-Spiking Neural Network for Energy-Efficient ECG Arrhythmia Classification

Key Points

Structured PICO

Main Result

Limitations

Implication

Abstract

Discussion

Cite This Study

Keep reading

Keep reading