Untersuchungen zur Etablierung eines EEG-Biomarkers im Rattenmodell der mesialen Temporallappenepilepsie mit Hippokampussklerose

Background: Epilepsy is a common neurological disorder, with mesial temporal lobe epilepsy with hippocampal sclerosis (mTLE-HS) being the most often occurring form of focal epilepsy in adults. mTLE-HS is often refractory to therapy and characterised by the occurrence of spontaneous epileptic seizures (SpES). These seizures can be recorded by electroencephalography (EEG). The occurrence of SpESs is a hallmark of an established epilepsy (EStE). An initial precipitating injury (IPI) such as head trauma, ischemia or hyperthermia can trigger the development of epilepsy. After an IPI, the development phase (EPG, epileptogenesis) begins. During and shortly after an IPI, acute symptomatic seizures (ASS EPG) may occurr. EPG cannot be diagnosed with certainty. Objective: To establish an EEG biomarker of EPG from ictal and non-ictal EEG in a rat model of mTLE-HS. Methods: In rats, electrodes for electrical stimulation were implanted into the perforant pathway and recording electrodes were placed into the dentate gyrus. After a 7 day recovery phase (BL, baseline), EPG was triggered by electrical stimulation of the perforant pathway (PPS) for 30 min on two consecutive days, followed by an 8 h stimulation on day three. EEG was continuously recorded during BL, EPG and EStE, capturing ictal EEG during BL (ASS BL), ASS EPG and SpESs as well as non-ictal EEG. Results: A subset of ictal EEG was segmented manually and descriptive features based on morphology were extracted. For intervals of both ictal and non-ictal EEG, features based on statistical properties were calculated. In 32 animals, ictal EEG (16.6% ASS BL, 58.3% ASS EPG, 25.1% SpES) was recorded. Non-ictal EEG was continuously gathered from 12 of these animals (63 d BL, 316 d EPG and 295d EstE in total). Modelling the segments of 49 representative ictal EEGs did not allow for better-than-chance discrimination between ASS and SpES segments. ASS EPG segments were correlated with spike and polyspike patterns of negative polarity. ASS BL segments were characterized by slow wave patterns, positive polarity and low amplitude. SpES segments were correlated with high amplitude patterns. After automated feature extraction and modelling with XGBoost, 88% of ASS EPG could be correctly identified. Most important features were coastline, intermittency and coherence. Correct classification of non-ictal EPG EEG intervals was successful in three animals (accuracy: 95.5 ± 2.7%). For another three animals, accuracy of correct EPG classification was high only at the beginning of the EPG phase (84.4 ± 8.7%). Features with highest importance were asymmetry and power in Theta 5-8 Hz, Gamma 100.1-160 Hz, Beta 14-30.9 and 20-30 Hz frequency bands. Conclusion: General suitability of ictal and non-ictal single-channel EEG for establishing a biomarker of EPG in an animal model of mTLE-HS was demonstrated and confirmed. However, performance of the utilised classification algorithms varied widely between subjects. In order to establish reliable EEG biomarkers, the presented methods should be extended and applied to larger data sets.