Electrocorticogram Analysis And Its Application In Epilepsy Surgical Treatment

Epilepsy is the most common brain disease after cerebrovascular diseases. It not only causes psychosomatic damage to patients, but also has great influence on their families and the society. Therefore, the diagnosis and treatment of epilepsy have always drawn attention from clinicians and neural scientists. To meet the practical need of electrocorticogram(ECoG) analysis in epilepsy surgeries, the present work investigates the seizure detection, epilepsy foci localization, functional brain mapping, and epilepsy mechanism from the neural informatics perspective. The work mainly consists of the following parts:First, the maximum entropy rate method based on recurrence plot is proposed to automatically detect epileptic seizures. The present method detects ictal period automatically by estimating information entropy of phase space distribution of multichannel signals. Research on inter-ictal and ictal signals with individual difference from several patients showed that the present method can differentiate inter-ictal and ictal signals correctly. Result suggests that it’s feasible for automatic seizure detection in clinical practice.Second, a spectral factorization-based current source density analysis method is proposed to conduct source analysis on epilepsy electrocorticogram signals to localize epileptic foci. Compared to traditional current source density methods, the spectral factorization based current source density methods could yield more accurate and stable results. In addition, it can cluster the results based on frequency features, thus better depict the discharge pattern of epileptic foci. This suggest that the spectral factorization-based current source density analysis method could detect seizure onsets automatically without ictal data.Third, a Monte Carlo based wavelet coherence(MC-WTC) method was developed to conduct time-frequency coherence analysis on ECoG data to build up cortical networks and define functional regions. In order to comply with the practical requirement of real-time analysis, large-scale parallel optimize design using GPGPU has been adopted. It turns out that MC-WTC method can efficiently mappping the high-level functional areas of the cortex, and the MC-WTC method using GPU acceleration can speed up the computation hugely, making it feasible to conduct real-time functional brain mapping in clinical environment.In addition, an epilepsy surgery assistance tool for clinician is developed. The key component of the toolbox is a 3-D brain visualization system. On account of the current surgery plan need from clinicians, flexible interface have been designed to perform brain model visualization, intracranial electrodes co-registration, brain functional plotting and brain network visualization, as well as to solve the positon offset problem in current electrode registration methods. The toolbox has friendly GUI and high efficiency, thus meet the time requirement of clinical practice. The toolbox can assist clinical doctors in surgery planning, surgical analysis, as well as surgery outcome evaluation.Finally, a harmonic wavelet transform based power law analysis method is proposed, and critical theory has been used to preliminarily investigate the mechanism of epilepsy development. Epilepsy development processes have been analysed based on ECoG data. The result showed that for neural system, the development of epilepsy is a dynamic process from an unstable state to a stable state, and the recovery of epilepsy is a process of recovery from a stable state to an unstable state, and prior to epilepsy onset, critical state of brain system alters. Critical theory analysis is a novel way to depict neural dynamics, and can be used to investigate epilepsy development, in addition, it can efficiently predict seizures.