Feature extraction with stacked autoencoders for epileptic seizure detection

Supratak, Akara; Li, Ling; Guo, Yike

doi:10.1109/embc.2014.6944546

Cited by 51 publications

(38 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…[15][16][17] In contrast to the process of hand-engineering features, deep learning (DL) methodologies learn intrinsic data features to obtain relevant data abstractions. 18 DL models have been used for seizure detection using EEG data streams, [19][20][21][22] in which impressive metric scores were achieved. This work aims to investigate for the first time the additional benefit that hemodynamic information derived from fNIRS recordings provides in a seizure detection task in the context of multimodal EEG-fNIRS recordings.…”

Section: Introductionmentioning

confidence: 99%

fNIRS improves seizure detection in multimodal EEG-fNIRS recordings

et al. 2019

View full text Add to dashboard Cite

In the context of epilepsy monitoring, electroencephalography (EEG) remains the modality of choice. Functional near-infrared spectroscopy (fNIRS) is a relatively innovative modality that cannot only characterize hemodynamic profiles of seizures but also allow for long-term recordings. We employ deep learning methods to investigate the benefits of integrating fNIRS measures for seizure detection. We designed a deep recurrent neural network with long short-term memory units and subsequently validated it using the CHBMIT scalp EEG database-a compendium of 896 h of surface EEG seizure recordings. After validating our network using EEG, fNIRS, and multimodal data comprising a corpus of 89 seizures from 40 refractory epileptic patients was used as model input to evaluate the integration of fNIRS measures. Following heuristic hyperparameter optimization, multimodal EEG-fNIRS data provide superior performance metrics (sensitivity and specificity of 89.7% and 95.5%, respectively) in a seizure detection task, with low generalization errors and loss. False detection rates are generally low, with 11.8% and 5.6% for EEG and multimodal data, respectively. Employing multimodal neuroimaging, particularly EEG-fNIRS, in epileptic patients, can enhance seizure detection performance. Furthermore, the neural network model proposed and characterized herein offers a promising framework for future multimodal investigations in seizure detection and prediction.

show abstract

Section: Introductionmentioning

confidence: 99%

fNIRS improves seizure detection in multimodal EEG-fNIRS recordings

et al. 2019

View full text Add to dashboard Cite

show abstract

“…Zabihi et al, [44] implemented their proposed method on the same dataset using MATLAB version R2014b with a 3.4 GHz processor and 16 GB RAM. The time required for feature extraction for one second of one EEG channel was reported as 2.6 ms. For an hour of EEG analysis on one channel, the time required for feature extraction with linear discriminant analysis (LDA) classification is about 9.6 s. Supratak et al [41] implemented their algorithm in MATLAB with a 3.4 GHz machine with 16.0 GB RAM. The reported training time varied from 2 to 5 h, depending on the amount of training for each patient, and the computation time for seizure detection was reported as~10 ms for each 1-second EEG segment.…”

Section: Discussionmentioning

confidence: 99%

Online EEG Seizure Detection and Localization

2019

View full text Add to dashboard Cite

Epilepsy is one of the three most prevalent neurological disorders. A significant proportion of patients suffering from epilepsy can be effectively treated if their seizures are detected in a timely manner. However, detection of most seizures requires the attention of trained neurologists—a scarce resource. Therefore, there is a need for an automatic seizure detection capability. A tunable non-patient-specific, non-seizure-specific method is proposed to detect the presence and locality of a seizure using electroencephalography (EEG) signals. This multifaceted computational approach is based on a network model of the brain and a distance metric based on the spectral profiles of EEG signals. This computationally time-efficient and cost-effective automated epileptic seizure detection algorithm has a median latency of 8 s, a median sensitivity of 83%, and a median false alarm rate of 2.9%. Hence, it is capable of being used in portable EEG devices to aid in the process of detecting and monitoring epileptic patients.

show abstract

“…Sensitivity of 83% is calculated over the whole experimental works. In [5] authors analyzed the CHB-MIT EEG data sets using stacked auto encoders. In this experimental work, the method detected 100% of Sensitivity, considering 6 numbers of patients.…”

Section: Revised Manuscript Received On February 05 2020mentioning

confidence: 99%

Automated Epileptic Seizure Detection Model using WPT, CFS and KNN based Multiclass TSVM

Mohapatra¹,

Mohanty²,

Swain³

et al. 2020

IJEAT

View full text Add to dashboard Cite

The main aim of the proposed work is to generate an accurate automated seizure detection model for the performance evaluation of the improvement on epileptic patients in an improved manner. Long data sets of EEG signals are recorded for a long duration of time which has taken from PhysioNet CHB-MIT EEG datset for this experimental work. Six types of elements are excerpted from EEG signals by using WPT method and which is then classified by using CFS method. Then, all the features are combinely inputted to the rule based twin- support vector machines (TSVMs ) to detect normal, ictal and pre-ictal EEG segments. The developed seizure detection WPT-KWMTSVM method achieved excellent performance with the average Accuracy, specificity, sensitivity, G-mean, positive predictive value, and Mathews correlation coefficients are 97.14%, 97.33%, 97.00%, 97.31%, 96.85%, 95.96% respectively The average area under curve (AUC) is approximately 1. The proposed method is able to enhance the seizure detection outcomes for proper clinical diagnosis in medical applications.

show abstract

Feature extraction with stacked autoencoders for epileptic seizure detection

Cited by 51 publications

References 12 publications

fNIRS improves seizure detection in multimodal EEG-fNIRS recordings

fNIRS improves seizure detection in multimodal EEG-fNIRS recordings

Online EEG Seizure Detection and Localization

Automated Epileptic Seizure Detection Model using WPT, CFS and KNN based Multiclass TSVM

Contact Info

Product

Resources

About