Dataset Information

Identifying signal-dependent information about the preictal state: A comparison across ECoG, EEG and EKG using deep learning.

ABSTRACT:

Background

The inability to reliably assess seizure risk is a major burden for epilepsy patients and prevents developing better treatments. Recent advances have paved the way for increasingly accurate seizure preictal state detection algorithms, primarily using electrocorticography (ECoG). To develop seizure forecasting for broad clinical and ambulatory use, however, less complex and invasive modalities are needed. Algorithms using scalp electroencephalography (EEG) and electrocardiography (EKG) have also achieved better than chance performance. But it remains unknown how much preictal information is in ECoG versus modalities amenable to everyday use - such as EKG and single channel EEG - and how to optimally extract that preictal information for seizure prediction.

Methods

We apply deep learning - a powerful method to extract information from complex data - on a large epilepsy data set containing multi-day, simultaneous recordings of EKG, ECoG, and EEG, using a variety of feature sets. We use the relative performance of our algorithms to compare the preictal information contained in each modality.

Results

We find that single-channel EKG contains a comparable amount of preictal information as scalp EEG with up to 21 channels and that preictal information is best extracted not with standard heart rate measures, but from the power spectral density. We report that preictal information is not preferentially contained in EEG or ECoG channels within the seizure onset zone.

Conclusion

Collectively, these insights may help to devise future prospective, minimally invasive long-term epilepsy monitoring trials with single-channel EKG as a particularly promising modality.

SUBMITTER: Meisel C

PROVIDER: S-EPMC6642360 | biostudies-literature | 2019 Jul

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Identifying signal-dependent information about the preictal state: A comparison across ECoG, EEG and EKG using deep learning.

Meisel Christian C Bailey Kimberlyn A KA

EBioMedicine 20190709

<h4>Background</h4>The inability to reliably assess seizure risk is a major burden for epilepsy patients and prevents developing better treatments. Recent advances have paved the way for increasingly accurate seizure preictal state detection algorithms, primarily using electrocorticography (ECoG). To develop seizure forecasting for broad clinical and ambulatory use, however, less complex and invasive modalities are needed. Algorithms using scalp electroencephalography (EEG) and electrocardiograp ...[more]

PMID: 31300348

Similar Datasets

Project description:Accurately characterising the brain networks involved in seizure activity may have important implications for our understanding of epilepsy. Intracranial EEG-fMRI can be used to capture focal epileptic events in humans with exquisite electrophysiological sensitivity and allows for identification of brain structures involved in this phenomenon over the entire brain. We investigated ictal BOLD networks using the simultaneous intracranial EEG-fMRI (icEEG-fMRI) in a 30 year-old male undergoing invasive presurgical evaluation with bilateral depth electrode implantations in amygdalae and hippocampi for refractory temporal lobe epilepsy. One spontaneous focal electrographic seizure was recorded. The aims of the data analysis were firstly to map BOLD changes related to the ictal activity identified on icEEG and secondly to compare different fMRI modelling approaches. Visual inspection of the icEEG showed an onset dominated by beta activity involving the right amygdala and hippocampus lasting 6.4 s (ictal onset phase), followed by gamma activity bilaterally lasting 14.8 s (late ictal phase). The fMRI data was analysed using SPM8 using two modelling approaches: firstly, purely based on the visually identified phases of the seizure and secondly, based on EEG spectral dynamics quantification. For the visual approach the two ictal phases were modelled as 'ON' blocks convolved with the haemodynamic response function; in addition the BOLD changes during the 30 s preceding the onset were modelled using a flexible basis set. For the quantitative fMRI modelling approach two models were evaluated: one consisting of the variations in beta and gamma bands power, thereby adding a quantitative element to the visually-derived models, and another based on principal components analysis of the entire spectrogram in attempt to reduce the bias associated with the visual appreciation of the icEEG. BOLD changes related to the visually defined ictal onset phase were revealed in the medial and lateral right temporal lobe. For the late ictal phase, the BOLD changes were remote from the SOZ and in deep brain areas (precuneus, posterior cingulate and others). The two quantitative models revealed BOLD changes involving the right hippocampus, amygdala and fusiform gyrus and in remote deep brain structures and the default mode network-related areas. In conclusion, icEEG-fMRI allowed us to reveal BOLD changes within and beyond the SOZ linked to very localised ictal fluctuations in beta and gamma activity measured in the amygdala and hippocampus. Furthermore, the BOLD changes within the SOZ structures were better captured by the quantitative models, highlighting the interest in considering seizure-related EEG fluctuations across the entire spectrum.

Project description:How and to what extent electrical brain activity reflects pharmacologically altered states and contents of consciousness, is not well understood. Therefore, we investigated whether measures of evoked and spontaneous electroencephalographic (EEG) signal diversity are altered by sub-anaesthetic levels of ketamine compared to normal wakefulness, and how these measures relate to subjective experience. High-density 62-channel EEG was used to record spontaneous brain activity and responses evoked by transcranial magnetic stimulation (TMS) in 10 healthy volunteers before and during administration of sub-anaesthetic doses of ketamine in an open-label within-subject design. Evoked signal diversity was assessed using the perturbational complexity index (PCI), calculated from EEG responses to TMS perturbations. Signal diversity of spontaneous EEG, with eyes open and eyes closed, was assessed by Lempel Ziv complexity (LZc), amplitude coalition entropy (ACE), and synchrony coalition entropy (SCE). Although no significant difference was found in TMS-evoked complexity (PCI) between the sub-anaesthetic ketamine condition and normal wakefulness, all measures of spontaneous EEG signal diversity (LZc, ACE, SCE) showed significantly increased values in the sub-anaesthetic ketamine condition. This increase in signal diversity correlated with subjective assessment of altered states of consciousness. Moreover, spontaneous signal diversity was significantly higher when participants had eyes open compared to eyes closed, both during normal wakefulness and during influence of sub-anaesthetic ketamine. The results suggest that PCI and spontaneous signal diversity may reflect distinct, complementary aspects of changes in brain properties related to altered states of consciousness: the brain's capacity for information integration, assessed by PCI, might be indicative of the brain's ability to sustain consciousness, while spontaneous complexity, as measured by EEG signal diversity, may be indicative of the complexity of conscious content. Thus, sub-anaesthetic ketamine may increase the complexity of the conscious content and the brain activity underlying it, while the level or general capacity for consciousness remains largely unaffected.

Dataset Information

Identifying signal-dependent information about the preictal state: A comparison across ECoG, EEG and EKG using deep learning.

Background

Methods

Results

Conclusion

Publications

Identifying signal-dependent information about the preictal state: A comparison across ECoG, EEG and EKG using deep learning.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets