Using Long Short-Term Memory (LSTM) recurrent neural networks to classify unprocessed EEG for seizure prediction DOI Creative Commons
Jordan D. Chambers, Mark Cook, Anthony N. Burkitt

et al.

Frontiers in Neuroscience, Journal Year: 2024, Volume and Issue: 18

Published: Nov. 15, 2024

Objective Seizure prediction could improve quality of life for patients through removing uncertainty and providing an opportunity acute treatments. Most seizure models use feature engineering to process the EEG recordings. Long-Short Term Memory (LSTM) neural networks are a recurrent network architecture that can display temporal dynamics and, therefore, potentially analyze signals without performing engineering. In this study, we tested if LSTMs classify unprocessed recordings make predictions. Methods Long-term intracranial data was used from 10 patients. 10-s segments were input LSTM trained signal. The final generated 5 outputs model over 50 s combined with time information account cycles. Results predictions significantly better than random predictor. When compared other publications using same dataset, our performed several others comparable best published date. Furthermore, framework still produce chance when experimental paradigm design altered, need reperform Significance Removing perform is advancement on previously models. This be applied many different patients’ needs variety interventions. Also, it opens possibility personalized altered meet daily needs.

Language: Английский

Concept-drifts adaptation for machine learning EEG epilepsy seizure prediction DOI Creative Commons

Edson David Pontes,

Mauro F. Pinto, Fábio Lopes

et al.

Scientific Reports, Journal Year: 2024, Volume and Issue: 14(1)

Published: April 8, 2024

Abstract Seizure prediction remains a challenge, with approximately 30% of patients unresponsive to conventional treatments. Addressing this issue is crucial for improving patients’ quality life, as timely intervention can mitigate the impact seizures. In research field, it critical identify preictal interval, transition from regular brain activity seizure. While previous studies have explored various Electroencephalogram (EEG) based methodologies prediction, few been clinically applicable. Recent underlined dynamic nature EEG data, characterised by data changes time, known concept drifts, highlighting need automated methods detect and adapt these changes. study, we investigate effectiveness automatic drift adaptation in seizure prediction. Three patient-specific approaches 10-minute horizon are compared: algorithm incorporating window adjustment method optimising performance Support Vector Machines (Backwards-Landmark Window), data-batch (seizures) selection using logistic regression (Seizure-batch Regression), integration classifiers (Dynamic Weighted Ensemble). These incorporate retraining process after each use combination univariate linear features SVM classifiers. The Firing Power was used post-processing technique generate alarms before were compared control approach on typical machine learning pipeline, considering group 37 Temporal Lobe Epilepsy EPILEPSIAE database. best-performing Window) achieved results 0.75 ± 0.33 sensitivity 1.03 1.00 false positive rate per hour. This new strategy performed above chance 89% surrogate predictor, whereas only validated 46%.

Language: Английский

Citations

3
Addressing data limitations in seizure prediction through transfer learning DOI Creative Commons
Fábio Lopes, Mauro F. Pinto, António Dourado

et al.

Scientific Reports, Journal Year: 2024, Volume and Issue: 14(1)

Published: June 19, 2024

Abstract According to the literature, seizure prediction models should be developed following a patient-specific approach. However, seizures are usually very rare events, meaning number of events that may used optimise approaches is limited. To overcome such constraint, we analysed possibility using data from patients an external database improve models. We present trained transfer learning procedure. deep convolutional autoencoder electroencephalogram 41 collected EPILEPSIAE database. Then, bidirectional long short-term memory and classifier layers were added on top encoder part optimised for 24 Universitätsklinikum Freiburg individually. The was as feature extraction module. Therefore, its weights not changed during training. Experimental results showed pretrained about four times fewer false alarms while maintaining same ability predict achieved more 13% validated patients. evidenced optimisation stable faster, saving computational resources. In summary, adopting represents significant advancement. It addresses limitation seen in field offers efficient training, conserving Additionally, despite compact size, allows easily share knowledge due ethical restrictions lower storage requirements. this study will shared with scientific community, promoting further research.

Language: Английский

Citations

3
Artificial intelligence in epilepsy phenotyping DOI
Andrew Knight, Tilo Gschwind, Peter D. Galer

et al.

Epilepsia, Journal Year: 2023, Volume and Issue: unknown

Published: Nov. 21, 2023

Artificial intelligence (AI) allows data analysis and integration at an unprecedented granularity scale. Here we review the technological advances, challenges, future perspectives of using AI for electro-clinical phenotyping animal models patients with epilepsy. In translational research, accurately identify behavioral states in epilepsy, allowing identification correlations between neural activity interictal ictal behavior. Clinical applications AI-based automated semi-automated audio video recordings people allow significant reduction reliable detection classification major motor seizures. can electrographic biomarkers such as spikes, high-frequency oscillations, seizure patterns. Integrating electroencephalographic, clinical, will contribute to optimizing therapy

Language: Английский

Citations

7
EpiNet: A Hybrid Machine Learning Model for Epileptic Seizure Prediction using EEG Signals from a 500 Patient Dataset DOI Open Access

Oishika Khair Esha,

Nasima Begum,

Shaila Rahman

et al.

International Journal of Advanced Computer Science and Applications, Journal Year: 2024, Volume and Issue: 15(1)

Published: Jan. 1, 2024

The accurate prognosis of epileptic seizures has great significance in enhancing the management epilepsy, necessitating creation robust and precise predictive models. EpiNet, our hybrid machine learning model for EEG signal analysis, incorporates key elements computer vision , positioning it within this advancing technological domain enhanced seizure prediction accuracy. Hence, research aims to provide a thorough investigation using Bonn Electroencephalogram (EEG) signals dataset as an alternative method. methodology used study encompasses training five models, such Support Vector Machines (SVM), Gaussian Naive Bayes, Gradient Boosting, XGBoost, LightGBM. Performance criteria, including accuracy, sensitivity, specificity, precision, recall, F1-score, are extensively assess efficacy each model. A unique contribution is development model, integrating predictions from individual models enhance overall accuracy epilepsy identification. Experimental results demonstrate notable success, with achieving 99.81%. matrices both classes model’s reliability. Visualizations, ROC-AUC curves curves, nuanced understanding models’ discriminative abilities performance improvement increasing sample size. comparative analysis existing studies reaffirms advancement research, at forefront prediction. This not only highlights promising integration medical diagnostics but also emphasises areas future refinement. achieved open avenues proactive healthcare improved patient outcomes.

Language: Английский

Citations

2
Review on the current long-term, limited lead electroencephalograms DOI
Adriana Ulate-Campos, Tobias Loddenkemper

Epilepsy & Behavior, Journal Year: 2023, Volume and Issue: 150, P. 109557 - 109557

Published: Dec. 8, 2023

Language: Английский

Citations

5
Epileptic seizure prediction based on multiresolution convolutional neural networks DOI Creative Commons

Ali K. Ibrahim,

Hanqi Zhuang, Emmanuelle Tognoli

et al.

Frontiers in Signal Processing, Journal Year: 2023, Volume and Issue: 3

Published: May 30, 2023

Epilepsy withholds patients’ control of their body or consciousness and puts them at risk in the course daily life. This article pursues development a smart neurocomputational technology to alert epileptic patients wearing EEG sensors an impending seizure. An innovative approach for seizure prediction has been proposed improve accuracy reduce false alarm rate comparison with state-of-the-art benchmarks. Maximal overlap discrete wavelet transform was used decompose signals into different frequency resolutions, multiresolution convolutional neural network is designed extract discriminative features from each band. The algorithm automatically generates patient-specific best classify preictal interictal segments subject. method can be applied any patient case dataset without need handcrafted feature extraction procedure. tested two popular epilepsy datasets. It achieved sensitivity 82% 0.058 Children’s Hospital Boston-MIT scalp 85% 0.19 American Society Seizure Prediction Challenge dataset. provides personalized solution that improved specificity, yet because algorithm’s intrinsic ability generalization, it emancipates reliance on epileptologists’ expertise tune wearable technological aid, which will ultimately help deploy broadly, including medically underserved locations across globe.

Language: Английский

Citations

4
Perceived seizure risk in epilepsy: Chronic electronic surveys with and without concurrent electroencephalography DOI Creative Commons
Jie Cui, Irena Balzekas, Ewan S. Nurse

et al.

Epilepsia, Journal Year: 2023, Volume and Issue: 64(9), P. 2421 - 2433

Published: June 12, 2023

Previous studies suggested that patients with epilepsy might be able to forecast their own seizures. This study aimed assess the relationships between premonitory symptoms, perceived seizure risk, and future recent self-reported electroencephalographically (EEG)-confirmed seizures in ambulatory natural home environments.

Language: Английский

Citations

4
Prospective validation of a seizure diary forecasting falls short DOI Creative Commons
Daniel M. Goldenholz,

Celena Eccleston,

Robert Moss

et al.

medRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: Jan. 13, 2024

Recently, a deep learning AI model forecasted seizure risk using retrospective diaries with higher accuracy than random forecasts. The present study sought to prospectively evaluate the same algorithm.

Language: Английский

Citations

1
Automated algorithms for seizure forecast: a systematic review and meta-analysis DOI Creative Commons
Ana Sofia Carmo, Mariana Abreu, Maria Fortuna Baptista

et al.

Journal of Neurology, Journal Year: 2024, Volume and Issue: unknown

Published: Sept. 6, 2024

Language: Английский

Citations

1
Multimodal wearable sensors inform cycles of seizure risk DOI Creative Commons
Nicholas M. Gregg, Tal Pal Attia, Mona Nasseri

et al.

medRxiv (Cold Spring Harbor Laboratory), Journal Year: 2022, Volume and Issue: unknown

Published: July 12, 2022

Abstract Objective Seizure unpredictability is a major source of disability for people with epilepsy. Recent work using chronic brain recordings has established that many individuals epilepsy seizure risk not random, but corresponds to circadian and multiday (multidien) cycles in excitability. Here, we aimed evaluate whether multimodal wearable device can characterize risk, compare wearables performance concurrent recordings. Methods Fourteen subjects underwent long-term ambulatory monitoring wrist worn (measuring heart rate, rate variability, accelerometry, tonic phasic electrodermal activity, temperature) an implanted responsive neurostimulation system interictal epileptiform abnormalities (IEA) electrographic seizures). Wavelet time-frequency analyses identified Circular statistics assessed phase locking physiology. Results Ten met inclusion criteria. The mean recording duration was 232 days. Seven had reliable detections (mean 76 occurred six (IEA), five (temperature), four (heart activity), three (accelerometry, activity) subjects. residual HR (HR after regression correlated physical activity (ACC)) increased Interpretation Long timescale cyclical changes are common epilepsy, seizures occur at preferred phases these individuals. Broadly accessible technology provide new insights into the chronobiology implications forecasting.

Language: Английский

Citations

3