Channel-annotated deep learning for enhanced interpretability in EEG-based seizure detection

Biomedical Signal Processing and Control, Год журнала: 2025, Номер 103, С. 107484 - 107484

Опубликована: Янв. 6, 2025

Язык: Английский

---

Understanding the Role of Self-Attention in a Transformer Model for the Discrimination of SCD From MCI Using Resting-State EEG

IEEE Journal of Biomedical and Health Informatics, Год журнала: 2024, Номер 28(6), С. 3422 - 3433

Опубликована: Апрель 18, 2024

The identification of EEG biomarkers to discriminate Subjective Cognitive Decline (SCD) from Mild Impairment (MCI) conditions is a complex task which requires great clinical effort and expertise. We exploit the self-attention component Transformer architecture obtain physiological explanations model's decisions in discrimination 56 SCD 45 MCI patients using resting-state EEG. Specifically, an interpretability workflow leveraging attention scores time-frequency analysis epochs through Continuous Wavelet Transform proposed. In classification framework, models are trained validated with 5-fold cross-validation evaluated on test set obtained by selecting 20% total subjects. Ablation studies hyperparameter tuning tests conducted identify optimal model configuration. Results show that best performing model, achieves acceptable results both epochs' patients' classification, capable finding specific patterns highlight changes brain activity between two conditions. demonstrate potential weights as tools guide experts understanding disease-relevant features could be discriminative MCI.

Язык: Английский

---

Automatic Detection of Abnormal EEG Signals Using WaveNet and LSTM

Sensors, Год журнала: 2023, Номер 23(13), С. 5960 - 5960

Опубликована: Июнь 27, 2023

Neurological disorders have an extreme impact on global health, affecting estimated one billion individuals worldwide. According to the World Health Organization (WHO), these neurological contribute approximately six million deaths annually, representing a significant burden. Early and accurate identification of brain pathological features in electroencephalogram (EEG) recordings is crucial for diagnosis management disorders. However, manual evaluation EEG not only time-consuming but also requires specialized skills. This problem exacerbated by scarcity trained neurologists healthcare sector, especially low- middle-income countries. These factors emphasize necessity automated diagnostic processes. With advancement machine learning algorithms, there great interest automating process early diagnoses using EEGs. Therefore, this paper presents novel deep model consisting two distinct paths, WaveNet-Long Short-Term Memory (LSTM) LSTM, automatic detection abnormal raw data. Through multiple ablation experiments, we demonstrated effectiveness importance all parts our proposed model. The performance was evaluated TUH Corpus V.2.0.0. (TUAB) achieved high classification accuracy 88.76%, which higher than existing state-of-the-art research studies. Moreover, generalization evaluating it another independent dataset, TUEP, without any hyperparameter tuning or adjustment. obtained 97.45% between normal recordings, confirming robustness

Язык: Английский

---

Unsupervised EEG-Based Seizure Anomaly Detection with Denoising Diffusion Probabilistic Models

International Journal of Neural Systems, Год журнала: 2024, Номер 34(09)

Опубликована: Май 17, 2024

While many seizure detection methods have demonstrated great accuracy, their training necessitates a substantial volume of labeled data. To address this issue, we propose novel method for unsupervised anomaly called SAnoDDPM, which uses denoising diffusion probabilistic models (DDPM). We designed pipeline that variable lower bound on Markov chains to identify potential values are unlikely occur in anomalous The model is first trained normal data, then data input the model. resamples and converts it Finally, presence seizures can be determined by comparing before after Moreover, 2D spectrograms encoded into vector-quantized representations, enables powerful efficient DDPM while maintaining its quality. Experimental comparisons publicly available datasets, CHB-MIT TUH, show our delivers better results, significantly reduces inference time, suitable deployment clinical environments. As far as aware, DDPM-based detection. This approach contributes progression algorithms, thereby augmenting practicality settings.

Язык: Английский

---

A Mutual Information-Based Many-Objective Optimization Method for EEG Channel Selection in the Epileptic Seizure Prediction Task

Cognitive Computation, Год журнала: 2024, Номер 16(3), С. 1268 - 1286

Опубликована: Март 23, 2024

Язык: Английский

---

End-to-end model for automatic seizure detection using supervised contrastive learning

Engineering Applications of Artificial Intelligence, Год журнала: 2024, Номер 133, С. 108665 - 108665

Опубликована: Май 28, 2024

Язык: Английский

---

ECn-MultiBSTM: multiclass epileptic seizure classification using electro cetacean optimized bidirectional long short-term memory model

Cognitive Neurodynamics, Год журнала: 2025, Номер 19(1)

Опубликована: Май 27, 2025

Язык: Английский

---

Variational Mode Decomposition-Based Moment Fusion for the Detection of Seizure Types From the Scalp EEG Measurements

IEEE Transactions on Instrumentation and Measurement, Год журнала: 2023, Номер 72, С. 1 - 12

Опубликована: Янв. 1, 2023

Accurate detection of seizure types is one the important research in field epilepsy for effective treatment and management. The identification potential biomarkers difficult challenging as nonlinear nonstationary characteristics electroencephalogram (EEG) vary within a type among types. In this work, novel variational mode decomposition (VMD) based feature fusion approach proposed to detect evolution seizures into electrographic seizures, complex partial focal bilateral tonic-clonic seizures. scalp EEGs these are decomposed using VMD. features, namely, zero crossing, Shannon entropy, approximate root squared zeroth moment (RSZM) extracted from bandlimited oscillatory modes. These features employed design three-class classification model random forest. order enhance performance further, RSZM measures adapted. framework experimented with ictal Temple University Hospital database. results indicate that six-level VMD performs better terms differentiating Among found effectively recognize first few seconds EEG. Further maximum accuracy 96.91% ten-fold cross-validation 85.2% patient-wise cross-validation. addition, our also be promising differentiation normal when we tested on wider datasets. Therefore, has great diagnosis.

Язык: Английский

---

Automated Multi-class Seizure-type Classification System using EEG Signals and Machine Learning Algorithms

IEEE Access, Год журнала: 2024, Номер 12, С. 136524 - 136541

Опубликована: Янв. 1, 2024

Язык: Английский

---

SMARTSeiz: Deep Learning With Attention Mechanism for Accurate Seizure Recognition in IoT Healthcare Devices

IEEE Journal of Biomedical and Health Informatics, Год журнала: 2023, Номер 28(7), С. 3810 - 3818

Опубликована: Дек. 6, 2023

The Internet of Things (IoT) is capable controlling the healthcare monitoring system for remote-based patients. Epilepsy, a chronic brain syndrome characterized by recurrent, unpredictable attacks, affects individuals all ages. IoT-based seizure can greatly enhance patients' quality life. IoT device acquires patient data and transmits it to computer program so that doctors examine it. Currently, invest significant manual effort in inspecting Electroencephalograph (EEG) signals identify activity. However, EEG-based detection algorithms face challenges real-world scenarios due non-stationary EEG variable patterns among patients recording sessions. Therefore, sophisticated computer-based approach necessary analyze complex records. In this work, authors proposed hybrid combining traditional convolution neural (CN) recurrent networks (RNN) along with an attention mechanism automatic recognition epileptic seizures through signal analysis. This focuses on subsets class recognition, resulting improved model performance. methods are evaluated using publicly available UCI dataset, which consists five classes: four normal conditions one abnormal condition. Experimental results demonstrate suggested achieves overall accuracy 97.05% five-class data, 99.52% binary classification distinguishing cases from instances. Furthermore, intelligent compatible IoMT (Internet Medical Things) cloud-based smart framework.

Язык: Английский

---