A multimodal imaging approach for quantitative assessment of epilepsy

Electroencephalography-functional magnetic resonance imaging (EEG-fMRI) coregistration and high density EEG (hdEEG) can be combined to noninvasively map abnormal brain activation elicited by epileptic processes. EEG-fMRI can provide information on the pathophysiological processes underlying interictal activity, since the hemodynamic changes are a consequence of the abnormal neural activity generating interictal epileptiform discharges (IEDs). The source analysis estimates the current density of the source that generates a measured electric potential and it yields a plausible dipole localization of irritative regions. The aim of this thesis is to develop a multimodal approach with hdEEG and EEG-fMRI coregistration in order to localize the epileptic activity and to verify the reliability of source localization and BOLD activation. In Chapter I the multimodal approach is introduced. The chapter is divided in two main parts: the first is based on EEG-fMRI coregistration and the second on the source localization in epilepsy. The first part consists of a brief review of the physiologic basis of EEG and fMRI and the technical basics of simultaneous recording, examining the conventional method for EEG-fMRI data. Technical challenges, safety issues, scanning modalities and EEG artifact removal strategies are also described. An overview of the state of EEG-fMRI is presented and the open problems of conventional analysis are discussed. The second part introduces the basic principles of the source estimation from EEG data in epilepsy and their limitations. The first chapter provides a general framework. The next two are devoted to introduce different approaches. Conventional analysis of EEG-fMRI data relies on spike-timing of epileptic activity: the neurologist identifies the intervals of the IEDs events, as represented by a square wave; this protocol is then convolved with a canonical hemodynamic response function (HRF) to construct a model for the general linear model (GLM) analysis. There are limitations to the technique, however. The conventional analysis is not automatic, suffers of subjectivity in IEDs classification, and using a suboptimal HRF to model the BOLD response the activation map may result over or under estimated. The novel method purposed integrates in the conventional GLM two new features: the regressor based on the EEG signal (Chapter II) and the individual-based hemodynamic response function (ibHRF) (Chapter III). In Chapter IV the performance of the novel method of EEG-fMRI data was tested on in silico data. Simulated fMRI datasets were created and used for the choice of the optimal HRF among five models: four standard and an individual-based HRF models. The performance of the method was evaluated using the Akaike information criterion as selection. Simulations would demonstrate the superiority of the novel method compared with the conventional ones and assess how the variations in HRF model affect the results of the statistical analysis. Chapter V introduces an automatic criterion aiming to separate in fMRI data the signal related to an internal network from the noise. After the decomposition process (probabilistic independent component analysis [PICA]), the optimal number of components was selected by applying a novel algorithm which takes into account, for each component, the mean values of the spatial activation maps followed by clustering, segmentation and spectral analysis steps. Comparing visual and automatic identification of the neuronal networks, the algorithm demonstrated high accuracy and precision. Thus, the automatic selection method allows to separate and detect the resting state networks reducing the subjectivity of the independent component assessment. In Chapter VI experimental design and analysis on real data are described. The chapter focuses on BOLD activity, source localization and agreement with the clinical history of twelve epileptic patients. The scope is to apply a multimodal approach combining noninvasive techniques of acquisition and analysis. Standard EEG and fMRI data were acquired during a single scanning session. The analysis of EEG-fMRI data was performed by using both the conventional GLM, the new GLM and the PICA. Source localization of IEDs was performed using 256-channels hdEEG. BOLD localizations were then compared to the EEG source reconstruction and to the expected epileptic activity defined by standard EEG and clinical outcome. The combination of multimodal techniques and their respectively methods of analysis are useful tools in the presurgical workup of epilepsy providing different methods of localization of the same epileptic foci. Furthermore, the combined use of EEG-fMRI and hdEEG offers a new and more complete approach to the study of epilepsy and may play an increasingly important role in the evaluation of patients with refractory focal epilepsy.