Application of a convolutional neural network for fully-automated detection of spike ripples in the scalp electroencephalogram

Nadalin, Jessica K.; Eden, Uri T.; Han, Xue; Richardson, R. Mark; Chu, Catherine J.; Kramer, Mark A.

doi:10.1016/j.jneumeth.2021.109239

Cited by 13 publications

(10 citation statements)

References 74 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Previously, we have shown good intra-rater reliability and performance using semi-automated 32 and fullyautomated 18,33 techniques on scalp EEG data. Here we introduce a combined detector that extracts both time series features and spectral information to detect spike ripples in intracranial data.…”

Section: Discussionmentioning

confidence: 90%

“…Our group has previously developed two separate spike ripple detectors for use in scalp EEG: a featurebased algorithm applied to time series data 18,32 , and a convolutional neural network (CNN) applied to spectrogram images 33 . To apply these tools to intracranial data, we first retrained the CNN detector on hand-marked events.…”

Section: Automated Spike Ripple Detectionmentioning

confidence: 99%

“…Finally, we randomly selected 1600, 1 s samples from the 5 channels outside of the EZ across all subjects as true negative detections. Spectrogram images of 1 s epochs (either containing a true positive, false positive, or true negative detection) were generated to train the CNN detector 33 . As in 33 , we applied transfer learning to train the CNN (i.e., we retrained layers of an existing neural network architecture using the spectrogram image training data) 34 .…”

Section: Automated Spike Ripple Detectionmentioning

confidence: 99%

“…Spectrogram images of 1 s epochs (either containing a true positive, false positive, or true negative detection) were generated to train the CNN detector 33 . As in 33 , we applied transfer learning to train the CNN (i.e., we retrained layers of an existing neural network architecture using the spectrogram image training data) 34 . The CNN returns the probability of a 1 s epoch containing a spike ripple.…”

Section: Automated Spike Ripple Detectionmentioning

confidence: 99%

“…detector that analyzes spectrogram images, which we previously developed 32,33 . Using a leave-one-out cross-validation procedure, we determined the probability threshold (0.65) of the CNN detector that achieved the highest precision of the combined detector against detections hand-marked by experts.…”

Section: The Intracranial Automated Spike Ripple Combined Detector Ha...mentioning

confidence: 99%

See 4 more Smart Citations

Spike ripples localize the epileptogenic zone better than other leading biomarkers: a multicenter intracranial study

Shi

Shaw

Walsh

et al. 2023

Preprint

Self Cite

View full text Add to dashboard Cite

ObjectiveWe evaluated whether the combination of epileptiform spikes and ripples (spike ripples) outperformed other leading biomarkers in identifying the epileptogenic zone across subjects in a multicenter international study.MethodsWe validated and applied an automated spike ripple detector on intracranial EEG recordings in subjects from 4 centers who subsequently underwent surgical resection with known 1-year seizure outcomes. We evaluated the spike ripple rate in subjects cured after resection (ILAE 1 outcome) and those with persistent seizures (ILAE 2-5) across sites and recording types. We also evaluated spike, wideband HFO (80-500 Hz), fast ripple (250-500 Hz), and ripple (80-250 Hz) rates using validated automated detectors. The proportion of resected events was computed and compared across subject outcomes and biomarkers.Results109 subjects were included. The majority of spike ripples were removed in subjects with ILAE 1 outcome (p = 1e-6), and this was qualitatively observed across the four sites (p = 0.032, p = 0.092, p = 0.0005, p = 0.003) and the two electrode types (p = 0.01, p = 7e-6). A higher proportion of spike ripples were removed in subjects with ILAE 1 outcomes compared to ILAE 2-5 outcomes (p = 0.02). Among ILAE 1 subjects, the proportion of spike ripples removed was higher than the proportion of spikes (p = 0.0004), wideband HFOs (p = 0.0004), fast ripples (p = 0.008), and ripples (p = 0.008) removed. At the individual level, more subjects with ILAE 1 outcome had the majority of spike ripples removed (40/48, 83%) than spikes (69%, p = 0.04), wideband HFOs (63%, p = 0.009), fast ripples (36%, p = 2e-5), or ripples (45%, p = 0.0007) removed.InterpretationWhen surgical resection was successful, the majority of spike ripples were removed. Automatically detected spike ripples have improved specificity for epileptogenic tissue compared to spikes, wideband HFOs, fast ripples, and ripples.

show abstract

Section: Discussionmentioning

confidence: 90%