Detection of hemodynamic responses to epileptic activity using simultaneous Electro-EncephaloGraphy (EEG)/Near Infra Red Spectroscopy (NIRS) acquisitions.

Simultaneous recordings of Electro-EncephaloGraphy (EEG) with Near InfraRed Spectroscopy (NIRS) allow measuring hemodynamic changes (changes in the concentration of oxy- and deoxyhemoglobin) at the time of epileptic discharges detected on scalp EEG. Two NIRS detection methods based on the General Linear Model (GLM) respectively in the time domain and in the time-frequency domain are investigated in this study using realistic simulations of spontaneous interictal epileptic activity. We evaluated the sensitivity at different Signal to Noise Ratios (SNR), the effect of either a large or a small number of discharges and the impact of model misspecification (e.g. omission or false detection of epileptic discharges). We also explored the effect on the estimation of key parameters, which set the model order. Simulations showed that both methods become inaccurate in lower SNR conditions, leading to many false positive detections. However, the time-frequency estimator showed better performance than the time-domain one. Key parameters for each algorithm were identified and results suggest to model confounds in the GLM differently for oxy- and deoxyhemoglobin. We also demonstrated that an inaccurate marking of epileptic events has a small impact on the detection statistics whereas an inaccurate specification of the hemodynamic response function delay decreases drastically the detection abilities. Finally, we illustrated the two methods on clinical EEG/NIRS data of one patient with focal epilepsy, showing an increase of regional Cerebral Blood Volume (rCBV) spatially concordant with the presumed epileptogenic focus.