Evaluation of a novel median power spectrogram for seizure detection by non-neurophysiologists

Yan, Paul; Melman, Tamar; Yan, Sherry; Otgonsuren, Munkhzul; Grinspan, Zachary M.

doi:10.1016/j.seizure.2017.06.016

Cited by 8 publications

(19 citation statements)

References 29 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…From CHB-MIT EEGs, 177 seizures and 177 randomly selected nonseizure epochs were converted to spectrograms. The EEGs were converted into spectrograms via the MPS method [17]. 90% of these seizure and non-seizure spectrograms were randomly selected to train and cross validate four different CNN models.…”

Section: Study Overviewmentioning

confidence: 99%

“…CHB-MIT EEGs were collected from 22 patients (ages 1. [5][6][7][8][9][10][11][12][13][14][15][16][17][18][19]. The waveforms were sampled at 256 Hz and included annotations of the seizure's start and end identified by neurophysiologists at CHB.…”

Section: Eeg Datamentioning

confidence: 99%

“…An alternative to the traditional CSA is the median power spectrogram (MPS) [17]. The MPS condenses the EEG into 1-3 spectrograms, and more clearly delineates the sloped harmonic banding pattern associated with the evolving rhythmic electrical activity that occur during electrographic seizures [18].…”

Section: Introductionmentioning

confidence: 99%

“…The MPS condenses the EEG into 1-3 spectrograms, and more clearly delineates the sloped harmonic banding pattern associated with the evolving rhythmic electrical activity that occur during electrographic seizures [18]. Seizures are more visually distinct on MPS, and clinicians need only brief (5 min) training to recognize them [17]. The brief training does not sacrifice performance, as physicians detect seizures on the MPS with comparable performance to studies using more complex qEEGs with longer training (65-86% sensitivity, 62-80% specificity) [13,14,[19][20][21].…”

Section: Introductionmentioning

confidence: 99%

“…A problem common to all qEEG interpretation is user inconsistency. Most spectrograms, including the MPS, demonstrate only moderate agreement among users (Cohen's Kappa 0.41-0.60) [14,17,20]. Furthermore, in typical clinical scenarios, the neurophysiologist and the bedside clinician cannot continuously monitor the qEEG.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Automated spectrographic seizure detection using convolutional neural networks

Yan

Wang

Kwok

et al. 2019

Seizure

Self Cite

View full text Add to dashboard Cite

Non-convulsive seizures are common in critically ill patients, and delays in diagnosis contribute to increased morbidity and mortality. Many intensive care units employ continuous EEG (cEEG) for seizure monitoring. Although cEEG is continuously recorded, it is often reviewed intermittently, which may delay seizure diagnosis and treatment. This may be mitigated with automated seizure detection. In this study, we develop and evaluate convolutional neural networks (CNN) to automate seizure detection on EEG spectrograms. Methods: Adult EEGs (12 patients, 12 EEGs, 33 seizures) from New-York Presbyterian Hospital (NYP) and pediatric EEGs (22 patients, 130 EEGs, 177 seizures) from Children's Hospital Boston (CHB) were converted into spectrograms. To simulate a telemetry display, seizure and non-seizure events on spectrograms were sequentially sampled as images across a detection window (26,380 total images). Four CNN models of increasing complexity (number of layers) were trained, cross-validated, and tested on CHB and NYP spectrographic images. All CNNs were based on the VGG-net architecture, with adjustments to alleviate overfitting. Results: For spectrographically visible seizures, two CNN models (containing 4 and 7 convolution layers) achieved > 90% seizure detection sensitivity and specificity on the CHB test set and > 90% sensitivity and 75-80% specificity on the NYP test set. The one CNN model (10 convolution layers) did not converge during training; while another CNN (2 convolution layers) performed poorly (60% sensitivity and 32% specificity) on the NYP test set. Conclusions: Seizure detection on EEG spectrograms with CNN models is feasible with sensitivity and specificity potentially suitable for clinical use.

show abstract

Section: Study Overviewmentioning

confidence: 99%