Phase-amplitude coupling between infraslow and high-frequency activities well discriminates between the preictal and interictal states

Biomimetic Deep Learning Networks With Applications to Epileptic Spasms and Seizure Prediction

OBJECTIVE

In this study, we present a novel biomimetic deep learning network for epileptic spasms and seizure prediction and compare its performance with state-of-the-art conventional machine learning models.

METHODS

Our proposed model incorporates modular Volterra kernel convolutional networks and bidirectional recurrent networks in combination with the phase amplitude cross-frequency coupling features derived from scalp EEG. They are applied to the standard CHB-MIT dataset containing focal epilepsy episodes as well as two other datasets from the Montefiore Medical Center and the University of California Los Angeles that provide data of patients experiencing infantile spasm (IS) syndrome.

RESULTS

Overall, in this study, the networks can produce accurate predictions (100%) and significant detection latencies (10 min). Furthermore, the biomimetic network outperforms conventional ones by producing no false positives.

SIGNIFICANCE

Biomimetic neural networks utilize extensive knowledge about processing and learning in the electrical networks of the brain. Predicting seizures in adults can improve their quality of life. Epileptic spasms in infants are part of a particular seizure type that needs identifying when suspicious behaviors are noticed in babies. Predicting epileptic spasms within a given time frame (the prediction horizon) suggests their existence and allows an epileptologist to flag an EEG trace for future review.

Pro-Ictal State in Human Temporal Lobe Epilepsy

BACKGROUND: Studies of continuous electroencephalography (EEG) suggest that seizures in individuals with focal-onset epilepsies preferentially occur during periods of heightened risk, typified by pathologic brain activities, termed pro-ictal states; however, the presence of (pathologic) pro-ictal states among a plethora of otherwise physiologic (e.g., sleep–wake cycle) states has not been established. METHODS: We studied a prospective, consecutive series of 15 patients with temporal lobe epilepsy who underwent limbic thalamic recordings in addition to routine (cortical) intracranial EEG for seizure localization. For each participant, pro-ictal (45 minutes before seizure onset) and interictal (4 hours removed from all seizures) EEG segments were divided into 10-minute, nonoverlapping windows, which were randomly distributed into training and validation cohorts in a 1:1 ratio. A deep neural classifier was applied to distinguish pro-ictal from interictal brain activities in a patient-specific fashion. RESULTS: We analyzed 1800 patient-hours of continuous thalamocortical EEG. Distinct pro-ictal states were detected in each participant. The median area under the receiver-operating characteristic curve of the classifier was 0.92 (interquartile range, 0.90–0.96). Pro-ictal states were distinguished at least 45 minutes before seizure onset in 13 of 15 participants; in 2 of 15 participants, they were distinguished up to 35 minutes prior. CONCLUSIONS: On the basis of thalamocortical EEG, pro-ictal states — pathologic brain activities during periods of heightened seizure risk — could be identified in patients with temporal lobe epilepsy and were detected, in our small sample, more than one half hour before seizure onset.

The enigma of infra-slow fluctuations in the human EEG

Spontaneous Infra-Slow Fluctuations (ISFs) of the human EEG (EEG-ISFs) were discovered 60 years ago when appropriate amplifiers for their recordings were designed. To avoid skin-related artifacts the recording of EEG-ISFs required puncturing the skin under the electrode. In the beginning of the 21st century the interest in EEG-ISFs was renewed with the appearance of commercially available DC-coupled amplified and by observation of ISFs of the blood oxygen level–dependent (BOLD) functional magnetic resonance imaging signal at a similar frequency. The independent components of irregular EEG-ISFs were shown to correlate with BOLD signals which in turn were driven by changes in arousal level measured by galvanic skin response (GSR), pupil size and HRV. There is no consensus regarding the temporal organization of EEG-ISFs: some studies emphasize the absence of peaks on EEG-ISFs spectra, some studies report prominent oscillations with frequency around 0.1 or 0.02 Hz, while some emphasize multiple discrete infraslow oscillations. No studies used parameters of EEG-ISFs as neuromarkers to discriminate psychiatric patients from healthy controls. Finally, a set of working hypotheses is suggested that must be tested in future research to solve the enigma of EEG-ISFs.

Pre- and post-synaptic A-type K+ channels regulate glutamatergic transmission and switch of the network into epileptiform oscillations

BACKGROUND AND PURPOSE

Anticonvulsants targeting K⁺ channels have not been clinically available, although neuronal hyperexcitability in seizures could be suppressed by activation of K⁺ channels. Voltage-gated A-type K⁺ channel (A-channel) inhibitors may be prescribed for diseases of neuromuscular junction but could cause seizures. Consistently, genetic loss of function of A-channels may also cause seizures. It is unclear why inhibition of A-channels, if compared with the other types of K⁺ channels, is particularly prone to seizure induction. This hinders the development of relevant therapeutic interventions.

EXPERIMENTAL APPROACH

The epileptogenic mechanisms of A-channel inhibition and antiepileptic actions of A-channel activation were investigated in electrophysiological and behavioral seizures with pharmacological and optogenetic maneuvers.

KEY RESULTS

Presynaptic Kv1.4 and postsynaptic Kv4.3 A-channels act synergistically to gate glutamatergic transmission and control rhythmogenesis in the amygdala. The interconnected neurons set into the oscillatory mode by A-channel inhibition would reverberate with regular paces and the same top frequency, demonstrating a spatiotemporally well-orchestrated system with built-in oscillatory rhythms normally curbed by A-channels. Accordingly, selective over-excitation of glutamatergic neurons or inhibition of A-channels suffices to induce behavioral seizures, which are effectively ameliorated by A-channel activators such as NS-5806 or AMPA receptor antagonists such as perampanel.

CONCLUSION AND IMPLICATIONS

Transsynaptic voltage-dependent A-channels serve as a biophysical-biochemical transducer responsible for a novel form of synaptic plasticity. Such a network-level switch into and out of the oscillatory mode may underlie a wide-scope of telencephalic information processing, or to its extreme, epileptic seizures. A-channels thus constitute a potential target of antiepileptic therapy.

Impact of DC-Coupled Electrophysiological Recordings for Translational Neuroscience: Case Study of Tracking Neural Dynamics in Rodent Models of Seizures

We propose that to fully understand biological mechanisms underlying pathological brain activity with transitions (e.g., into and out of seizures), wide-bandwidth electrophysiological recordings are important. We demonstrate the importance of ultraslow potential shifts and infraslow oscillations for reliable tracking of synaptic physiology, within a neural mass model, from brain recordings that undergo pathological phase transitions. We use wide-bandwidth data (direct current (DC) to high-frequency activity), recorded using epidural and penetrating graphene micro-transistor arrays in a rodent model of acute seizures. Using this technological approach, we capture the dynamics of infraslow changes that contribute to seizure initiation (active pre-seizure DC shifts) and progression (passive DC shifts). By employing a continuous-discrete unscented Kalman filter, we track biological mechanisms from full-bandwidth data with and without active pre-seizure DC shifts during paroxysmal transitions. We then apply the same methodological approach for tracking the same parameters after application of high-pass-filtering >0.3Hz to both data sets. This approach reveals that ultraslow potential shifts play a fundamental role in the transition to seizure, and the use of high-pass-filtered data results in the loss of key information in regard to seizure onset and termination dynamics.