Related Experiment Video
Updated: Jun 13, 2026
A Pipeline for 3D Multimodality Image Integration and Computer-assisted Planning in Epilepsy Surgery
Published on: May 20, 2016
Pyramid Vision Transformer-Enhanced Conformer Network for Epileptic Seizure Recognition Using MultiChannel EEG
Weiguang Dong1, Jian Lian2, Xinyu Wang1
1School of Transportation and Vehicle Engineering, Wuxi University, Wuxi 214105, P. R. China.
None:
The classification of electroencephalogram (EEG) signals plays an important role in neuroscience research and clinical diagnosis of epileptic seizures. This work aims to solve EEG data classification tasks by using the Children's Hospital Boston and Massachusetts Institute of Technology (CHB-MIT) dataset, and proposes an improved Conformer structure strengthened with the Pyramid Vision Transformer (PVT). The major purpose of this study is to obtain higher classification precision and extract more distinguishable features from EEG signals. An enhanced Conformer-based framework is presented for EEG classification tasks. The PVT-Enhanced Conformer integrates the advantages of Convolutional Neural Networks (CNNs) and PVT. Specifically, the PVT module is inserted into the Conformer to optimize the original self-attention mechanism. In addition, the model is trained with cross-entropy as the loss function and the Adaptive Momentum (Adam) optimization algorithm for parameter updating. Experimental results on the CHB-MIT dataset show that the proposed model performs better than existing state-of-the-art models in three key indicators: classification accuracy (99.23%), specificity (99.61%), and sensitivity (98.11%). Furthermore, ablation experiments are conducted to evaluate the contribution of each component, which verifies the effectiveness of both PVT integration and the feature fusion strategy adopted in this work. This research provides an improved Conformer architecture for EEG data classification based on the CHB-MIT dataset. The results show the potential of the proposed method, which is expected to promote the development of artificial intelligence and biomedical engineering cases. Cross-dataset validation on Bonn and Siena datasets further confirms the method's generalization capability. Future work may focus on further refining the proposed approach and exploring its application in the diagnosis of other neurological diseases.
