Post-ictal generalized EEG suppression and seizure-induced mortality are reduced by enhancing dorsal raphe serotonergic neurotransmission

Abstract

<jats:title>Abstract</jats:title><jats:p>Sudden unexpected death in epilepsy (SUDEP) is the leading cause of death in patients with refractory epilepsy. A proposed risk marker for SUDEP is the duration of post-ictal generalized EEG suppression (PGES). The mechanisms underlying PGES are unknown. Serotonin (5-HT) has been implicated in SUDEP pathophysiology. Seizures suppress activity of 5-HT neurons in the dorsal raphe nucleus (DRN). We hypothesized that suppression of DRN 5-HT neuron activity contributes to PGES and increasing 5-HT neurotransmission or stimulating the DRN before a seizure would decrease PGES duration. Adult C57BL/6 and <jats:italic>Pet1-Cre</jats:italic> mice received EEG/EMG electrodes, a bipolar stimulating/recording electrode in the right basolateral amygdala, and either a microdialysis guide cannula or an injection of adeno-associated virus (AAV) allowing expression of channelrhodopsin2 plus an optic fiber into the DRN. Systemic application of the selective 5-HT reuptake inhibitor citalopram (20 mg/kg) decreased PGES duration from seizures induced during wake (n = 23) and NREM sleep (n = 13) whereas fluoxetine (20 mg/kg) pretreatment decreased PGES duration following seizures induced from wake (n = 11), but not NREM sleep (n = 9). Focal chemical (n = 6) or optogenetic (n = 8) stimulation of the DRN reduced PGES duration following kindled seizures and reduced morality following maximal electroshock seizures (n = 6) induced during wake. During PGES, animals exhibited immobility and suppression of EEG activity that was reduced by citalopram pretreatment. These results indicate that 5-HT and the DRN may regulate PGES and seizure-induced mortality.</jats:p><jats:sec><jats:title>Highlights</jats:title><jats:list list-type="simple"><jats:list-item><jats:label>-</jats:label><jats:p>PGES consistently follows seizures induced by amygdala stimulation in amygdala-kindled mice.</jats:p></jats:list-item><jats:list-item><jats:label>-</jats:label><jats:p>Seizure-induced dysregulation of 5-HT neurotransmission from the dorsal raphe nucleus may contribute to PGES.</jats:p></jats:list-item><jats:list-item><jats:label>-</jats:label><jats:p>Systemic administration of 5-HT enhancing drugs and stimulation of the DRN reduces PGES duration.</jats:p></jats:list-item><jats:list-item><jats:label>-</jats:label><jats:p>PGES is associated with post-ictal immobility in kindled mice that can be reduced by pretreatment with citalopram.</jats:p></jats:list-item><jats:list-item><jats:label>-</jats:label><jats:p>Recovery of EEG frequencies to baseline occurs in a stepwise manner with the lowest frequencies recovering first.</jats:p></jats:list-item></jats:list></jats:sec>

A hybrid unsupervised and supervised learning approach for postictal generalized EEG suppression detection

“…The mean feature measures a probability distribution's central tendency. The kurtosis and skewness features measure the tailedness and the asymmetry of a probability distribution, respectively (Li et al, 2020). Hjorth parameters are generally used in feature extraction for EEG signal analysis (Charbonnier et al, 2011) including activity, mobility, and complexity (Redmond and Heneghan, 2006).…”

“…Based on the definition of PGES, it seems straightforward to identify a period of low-amplitude EEG signals (< 10 µV). However, real-world data recorded in EMUs may contain high-amplitude signals due to physiological artifacts such as respiration, muscle, and movement-related artifacts (Li et al, 2020). Therefore, in practice the duration of PGES is determined manually with visual inspection of EEG signal readings by clinical experts, who can leverage additional video recordings along with signals to identify high-amplitude artifacts that are not real EEG activities (Theeranaew et al, 2017).…”

“…However, such a manual task is time-consuming and labor-intensive, and the judging criteria of PGES with artifacts by each clinical expert are not standardized, which may be subjective and unreliable (Zhao et al, 2021). Automatic PGES detection tools are highly desirable to help clinical experts review and annotate PGES in EEG recordings (Li et al, 2020).…”

“…Automated techniques have been studied for PGES detection, including a logistic regression approach based on frequency-domain features (Theeranaew et al, 2017), an eXtreme Gradient Boosting (XGBoost) classifier with timedomain and entropy-based features (Mier et al, 2020), and deep learning models based on convolutional neural network Vance et al, 2020). However, these studies utilized segment-based evaluation (i.e., the predictions for each segment determine the performance metrics), which has been demonstrated to be ineffective in measuring the performance of PGES detection in real clinical settings (Li et al, 2020). In a previous study, we introduced a more practically relevant manner (known as recording-based evaluation) to evaluate automated PGES detection methods based on the time distance, which is the time difference between the detected PGES end time and the actual expert-annotated end time (Li et al, 2020).…”

“…However, these studies utilized segment-based evaluation (i.e., the predictions for each segment determine the performance metrics), which has been demonstrated to be ineffective in measuring the performance of PGES detection in real clinical settings (Li et al, 2020). In a previous study, we introduced a more practically relevant manner (known as recording-based evaluation) to evaluate automated PGES detection methods based on the time distance, which is the time difference between the detected PGES end time and the actual expert-annotated end time (Li et al, 2020). With such time distance-based evaluation metrics, we developed and evaluated a feature-based random forest (RF) approach for automatic PGES detection with multi-channel EEG recordings.…”