Locally-adaptive feature selection for nonconvulsive seizure detection

Biomedical Signal Processing and Control, Journal Year: 2025, Volume and Issue: 105, P. 107535 - 107535

Published: Jan. 31, 2025

Language: Английский

---

Transformers in EEG Analysis: A Review of Architectures and Applications in Motor Imagery, Seizure, and Emotion Classification

Sensors, Journal Year: 2025, Volume and Issue: 25(5), P. 1293 - 1293

Published: Feb. 20, 2025

Transformers have rapidly influenced research across various domains. With their superior capability to encode long sequences, they demonstrated exceptional performance, outperforming existing machine learning methods. There has been a rapid increase in the development of transformer-based models for EEG analysis. The high volumes recently published papers highlight need further studies exploring transformer architectures, key components, and employed particularly studies. This paper aims explore four major architectures: Time Series Transformer, Vision Graph Attention hybrid models, along with variants recent We categorize according most frequent applications motor imagery classification, emotion recognition, seizure detection. also highlights challenges applying transformers datasets reviews data augmentation transfer as potential solutions explored years. Finally, we provide summarized comparison reported results. hope this serves roadmap researchers interested employing architectures

Language: Английский

---

EEG-Based Seizure Prediction Using Hybrid DenseNet–ViT Network with Attention Fusion

Brain Sciences, Journal Year: 2024, Volume and Issue: 14(8), P. 839 - 839

Published: Aug. 21, 2024

Epilepsy seizure prediction is vital for enhancing the quality of life individuals with epilepsy. In this study, we introduce a novel hybrid deep learning architecture, merging DenseNet and Vision Transformer (ViT) an attention fusion layer prediction. captures hierarchical features ensures efficient parameter usage, while ViT offers self-attention mechanisms global feature representation. The effectively amalgamates from both networks, guaranteeing most relevant information harnessed raw EEG signals were preprocessed using short-time Fourier transform (STFT) to implement time-frequency analysis convert into matrices. Then, they fed proposed DenseNet-ViT network model achieve end-to-end CHB-MIT dataset, including data 24 patients, was used evaluation leave-one-out cross-validation method utilized evaluate performance model. Our results demonstrate superior in prediction, exhibiting high accuracy low redundancy, which suggests that combining DenseNet, ViT, mechanism can significantly enhance capabilities facilitate more precise therapeutic interventions.

Language: Английский

---

Tiny CNN-Based Supervised Contrastive Learning for Seizure Prediction

International Journal of Neural Systems, Journal Year: 2025, Volume and Issue: unknown

Published: April 5, 2025

Automatic seizure prediction based on ElectroEncephaloGraphy (EEG) ensures the safety of patients with epilepsy and mitigates anxiety. In recent years, significant progress has been made in this field. However, predictive performance existing methods encounters a bottleneck that is difficult to overcome. Moreover, there are certain limitations such as differences efficacy among or intricate model structures. Given these considerations, Siamese Network (SiaNet) Triplet (TriNet) proposed tiny convolutional neural network supervised contrastive learning. Short-Time Fourier Transform (STFT) first applied pre-processed data. Then data tuples constructed fed into networks for training. Both try minimize interval between samples same class while maximize different classes. The two consist multiple branches shared weights, which can learn from each other via Promising results obtained CHB-MIT Siena datasets, total 35 patients. Meanwhile, both models have only 19.351K parameters.

Language: Английский

---

A Lightweight Convolutional Neural Network-Reformer Model for Efficient Epileptic Seizure Detection

International Journal of Neural Systems, Journal Year: 2024, Volume and Issue: 34(12)

Published: Aug. 30, 2024

A real-time and reliable automatic detection system for epileptic seizures holds significant value in assisting physicians with rapid diagnosis treatment of epilepsy. Aiming to address this issue, a novel lightweight model called Convolutional Neural Network-Reformer (CNN-Reformer) is proposed seizure on long-term EEG. The CNN-Reformer consists two main parts: the Data Reshaping (DR) module Efficient Attention Concentration (EAC) module. This framework reduces network parameters while retaining effective feature extraction multi-channel EEGs, thereby improving computational efficiency performance. Initially, raw EEG signals undergo Discrete Wavelet Transform (DWT) signal filtering, then fed into DR data compression reshaping preserving local features. Subsequently, these features are sent EAC extract global perform categorization. Post-processing involving sliding window averaging, thresholding, collar techniques further deployed reduce false rate (FDR) improve On CHB-MIT scalp dataset, our method achieves an average sensitivity 97.57%, accuracy 98.09%, specificity 98.11% at segment-based level, 96.81%, along FDR 0.27/h, latency 17.81 s event-based level. SH-SDU dataset we collected, yielded 94.51%, 92.83%, 92.81%, 94.11%. testing time 1[Formula: see text]h 1.92[Formula: text]s. excellent results fast speed demonstrate its potential efficient detection.

Language: Английский

---

CNN-Informer: A Hybrid Deep Learning Model for Seizure Detection on Long-term EEG

Neural Networks, Journal Year: 2024, Volume and Issue: 181, P. 106855 - 106855

Published: Oct. 28, 2024

Language: Английский

---

A new fuzzy-based ensemble framework based on attention-based deep learning architectures for automated detection of abnormal EEG

International Journal of Systems Assurance Engineering and Management, Journal Year: 2024, Volume and Issue: 15(12), P. 5713 - 5725

Published: Nov. 7, 2024

Language: Английский

---