LMA-EEGNet: A Lightweight Multi-Attention Network for Neonatal Seizure Detection Using EEG signals

Weicheng Zhou,Wei Zheng,Youbing Feng,Xiaolong Li

DOI: https://doi.org/10.3390/electronics13122354

IF: 2.9

2024-06-17

Electronics

Abstract:Neonatal epilepsy is an early postnatal brain disorder, and automatic seizure detection is crucial for timely diagnosis and treatment to reduce potential brain damage. This work proposes a novel Lightweight Multi-Attention Network, LMA-EEGNet, for diagnosing neonatal epileptic seizures from multi-channel EEG signals employing dilated depthwise separable convolution (DDS Conv) for feature extraction and using pointwise convolution followed by global average pooling for classification. The proposed approach substantially reduces the model size, number of parameters, and computational complexity, which are crucial for real-time detection and clinical diagnosis of neonatal epileptic seizures. LMA-EEGNet integrates temporal and spectral features through distinct temporal and spectral branches. The temporal branch uses DDS Conv to extract temporal features, enhanced by a channel attention mechanism. The spectral branch utilizes similar convolutions alongside a spatial attention mechanism to highlight key frequency components. Outputs from both branches are merged and processed through a pointwise convolution layer and a global average pooling layer for efficient neonatal seizure detection. Experimental results show that our model, with only 2471 parameters and a size of 23 KB, achieves an accuracy of 95.71% and an AUC of 0.9862, demonstrating its potential for practical deployment. This study provides an effective deep learning solution for the early detection of neonatal epileptic seizures, improving diagnostic accuracy and timeliness.

engineering, electrical & electronic,computer science, information systems,physics, applied

What problem does this paper attempt to address?

### Problems the Paper Aims to Solve

The paper aims to address the issue of automatic detection of neonatal seizures. Specifically, it proposes a lightweight multi-attention network (LMA-EEGNet) for diagnosing neonatal seizures through multi-channel electroencephalogram (EEG) signals. The main objectives include:

1. **Improving Detection Accuracy**: Enhance the accuracy of neonatal seizure detection by utilizing a lightweight network architecture and multi-attention mechanisms.
2. **Reducing False Positive Rate**: Lower the false alarm rate by optimizing feature extraction methods.
3. **Reducing Computational Complexity**: Significantly reduce model size, the number of parameters, and computational complexity by using dilated depthwise separable convolution (DDS Conv), making it suitable for real-time monitoring and clinical diagnosis.
4. **Integrating Time-Frequency Features**: Combine temporal and spectral branches to extract temporal and spectral features of EEG signals, respectively, and enhance critical information through attention mechanisms.

Through these innovations, LMA-EEGNet provides an efficient and practical solution for the early detection of neonatal seizures, improving diagnostic accuracy and timeliness.