Independent component analysis as a model‐free approach for the detection of BOLD changes related to epileptic spikes: A simulation study

Abstract

EEG-fMRI in epileptic patients is commonly analyzed using the general linear model (GLM), which assumes a known hemodynamic response function (HRF) to epileptic spikes in the EEG. In contrast, independent component analysis (ICA) can extract Blood-Oxygenation Level Dependent (BOLD) responses without imposing constraints on the HRF. This technique was evaluated on data generated by superimposing artificial responses on real background fMRI signals. Simulations were run using a wide range of EEG spiking rates, HRF amplitudes, and activation regions. The data were decomposed by spatial ICA into independent components. A deconvolution method then identified component time courses significantly related to the simulated spikes, without constraining the shape of the HRF. Components matching the simulated activation regions ("concordant components") were found in 84.4% of simulations, while components at discordant locations were found in 12.2% of simulations. These false activations were often related to large artifacts that coincidentally occurred simultaneously with some of the random simulated spikes. The performance of the method depended closely on the simulation parameters; when the number of spikes was low, concordant components could only be identified when HRF amplitudes were large. Although ICA did not depend on the shape of the HRF, data processed with the GLM did not reveal the appropriate activation region when the HRF varied slightly from the canonical shape used in the model. ICA may thus be able to extract BOLD responses from EEG-fMRI data in epileptic patients, in a way that is robust to uncertainty and variability in the shape of the HRF.

Modulation by EEG features of BOLD responses to interictal epileptiform discharges

“…This involved correcting for slice timing differences using sinc interpolation, high-pass filtering to remove scanner drift, masking to remove voxels outside the brain (Smith, 2002), and calculating the independent components by an iterative fixed-point method maximizing the non-Gaussianity of the sources, which has been shown to be equivalent to maximizing statistical independence (Hyvarinen and Oja, 2000). The number of extracted sources was determined by performing 20 repetitions of the ICA decomposition with random initialization, as true sources are likely to appear consistently in multiple ICA decompositions (Himberg et al, 2004;LeVan and Gotman, 2009). The reproducible components were then identified using the RAICAR algorithm (Yang et al, 2008), which matches components across ICA realizations based on their pairwise spatial cross-correlation coefficient.…”

“…In contrast, the ICA method allows for arbitrary HRFs. It is especially well-suited as an exploratory analysis to characterize the BOLD response from the time course of IED-related components (Rodionov et al, 2007;LeVan and Gotman, 2009). It is only after characterizing the relationship between IEDs and HRFs that it is then possible to integrate it into the model.…”

“…This exploratory method can extract consistent regions of activations and associated time courses based on a criterion of spatial independence, with minimal prior assumptions on the data. Careful examination of the potentially large number of extracted components may reveal regions of activation related to the epileptic discharges (Rodionov et al, 2007;LeVan and Gotman, 2009). ICA does not impose any constraints on the HRF; rather, an analysis of the component time courses may yield information about the BOLD response that had not been initially hypothesized.…”

Independent component analysis reveals dynamic ictal BOLD responses in EEG-fMRI data from focal epilepsy patients

“…The independent components were then computed by an iterative fixed-point method maximizing the non-Gaussianity of the sources, which has been shown to be equivalent to maximizing statistical independence (Hyvarinen and Oja, 2000). The number of extracted sources was determined by performing 20 repetitions of the ICA decomposition with random initialization, as true sources are likely to appear consistently in multiple ICA decompositions (Himberg et al, 2004;LeVan and Gotman, 2009). The reproducible components were then identified using the RAICAR algorithm (Yang et al, 2008), which matches components across ICA realizations based on their pairwise spatial cross-correlation coefficient.…”

EEG-fMRI Multimodal Integration for Epilepsy Research

“…Moreover, the classifier was trained on GLM-activation on a very small number of healthy subjects and the training was implemented on only 7% of the testing dataset, decreasing the statistical significance of the model. LeVan and Gotman, 2009) introduced a more independent ICA method using deconvolution for identifying component time courses significantly related to simulated focal spikes without constraining the shape of the HRF. Artificial time courses were obtained by generating spikes at random tims and convolving them with a canonical HRF computed from the difference of two gamma functions (Glover, 1999), and varying the location of the activation, the number of simulated spikes per run, and the HRF amplitude.…”