Study of the Hemodynamic Response to Interictal Epileptiform Discharges in Human Epilepsy Using Functional Near Infrared Spectroscopy

Functional near-infrared spectroscopy (fNIRS) has emerged as a promising neuroimaging technique as it allows non-invasive and long-term monitoring of cortical hemodynamics. For the last decades, fNIRS combined with electroencephalography (EEG) has been applied in the context of human epilepsy, and has yielded good results. However, most previous work only focused on the study of epileptic seizures which are random and seldom occur during EEG-fNIRS testing. This thesis sought to evaluate the potential of EEG-fNIRS in observing the hemodynamic changes associated with interictal epileptiform discharges (IEDs), and to determine whether these IEDs can also be used to extract useful information in the localization of the epileptic focus site. Based on the EEG-fNIRS data collected from a relatively large number of patients (40) and using a standard general linear model (GLM) approach, the first study of this thesis provided preliminary estimates of the sensitivity and the specificity of EEG-fNIRS in detecting brain areas activated by IEDs and in localizing the epileptic focus region. In the 29 patients with neocortical epilepsies, significant deoxygenated hemoglobin (HbR) concentration decreases and oxygenated hemoglobin (HbO) concentration increases corresponding to IEDs were observed in 62% and 38% of patients respectively. This HbR/HbO response was most significant in the epileptic focus region among all the activations, and thus could lead to successful identification of the epileptic focus site in 28%/21% of the patients. These modest estimates of the sensitivity and the specificity suggested that using a standard GLM with a canonical hemodynamic response function (HRF) might not be the optimal method in the analysis of IEDs. Therefore, the second study of this thesis made a first attempt to model the patient-specific variations in the shape of the hemodynamic response to IEDs. A quadratic term was also added to the model to account for the nonlinearity in the response when frequent IEDs were present in the recording. The new models were first validated through carefully designed simulations, and were then applied in the data analysis of five selected patients. Compared with the canonical HRF, including patient-specific HRFs in the GLM analysis not only significantly improved the statistical scores and the spatial extents of existing activations, but also was able to detect new brain regions activated by IEDs on all of the five patients. These improvements in activation detection also helped obtain more accurate focus localization results in some cases.