Investigating the effect of template head models on Event-Related Potential source localization: a simulation and real-data study

Abstract

Introduction: Event-Related Potentials (ERPs) are valuable for studying brain activity with millisecond-level temporal resolution. While the temporal resolution of this technique is excellent, the spatial resolution is limited. Source localization aims to identify the brain regions generating the EEG data, thus increasing the spatial resolution, but its accuracy depends heavily on the head model used. This study compares the performance of subject-specific and template-based head models in both simulated and real-world ERP localization tasks.

Methods: Simulated data mimicking realistic ERPs was created to evaluate the impact of head model choice systematically, after which subject-specific and template-based head models were used for the reconstruction of the data. The different modeling approaches were also applied to a face recognition dataset.

Results: The results indicate that the template models capture the simulated activity less accurately, producing more spurious sources and identifying less true sources correctly. Furthermore, the results show that while creating more accurate and detailed head models is beneficial for the localization accuracy when using subject-specific head models, this is less the case for template head models. The main N170 source of the face recognition dataset was correctly localized to the fusiform gyrus, a known face processing area, using the subject-specific models. Apart from the fusiform gyrus, the template models also reconstructed several other sources, illustrating the localization inaccuracies.

Discussion: While template models allow researchers to investigate the neural generators of ERP components when no subject-specific MRIs are available, it could lead to misinterpretations. Therefore, it is important to consider a priori knowledge and hypotheses when interpreting results obtained with template head models, acknowledging potential localization errors.

Keywords: EEG; EEG source localization; ERP; Event-Related Potentials; head modeling; source estimation.

Figure 1

Illustration of the calculation of sensitivity and precision of the source reconstructions for the simulated data. True positives (TPs) are defined as reconstructed clusters within a 3 cm distance from the center of the simulated ROIs, while false negatives (FNs) are ROIs without a nearby cluster, and false positives (FPs) are clusters not within 3 cm of any ROI. Sensitivity is calculated as the ratio of TPs to the sum of TPs and FNs, and precision is calculated as the ratio of TPs to the sum of TPs and FPs.

Figure 2

Overview of the simulated data at sensor level averaged over all subjects. The simulated epochs in the ERP condition at SNR = -10 dB are averaged.

Figure 3

Overview of the original simulated data and the reconstructed activity averaged over all subjects for the different networks at SNR = -10 dB. For the reconstructed activity, the difference between the ERP- and the noise-conditions is shown. In the case of the simulations and the subject-specific reconstructions, the source activity was morphed to the anatomy of the average head model before averaging.

Figure 4

Results of the quantification of the localization errors for the SNR = -10 dB. In this evaluation the sensitivity and the precision of the obtained sources, the localization error and the spatial dispersion of these reconstructed sources were taken into account. For each of these measures, the difference between using the subject-specific and average head models is shown for each of the different modeling approaches. Clusters of activity were considered to be correctly localized when the difference between the center of the reconstructed cluster was within 3 cm of the center of the simulated ROIs.

Figure 5

Results of the quantification of the localization errors. In this evaluation the sensitivity and the precision of the obtained sources, the localization error and the spatial dispersion of these reconstructed sources were taken into account. For each of these measures, the effect of both the SNR of the simulated data and the difference between using the subject-specific and average head models is shown for each of the different modeling approaches. Clusters of activity were considered to be correctly localized when the difference between the center of the reconstructed cluster was within 3 cm of the center of the simulated ROIs.

Figure 6

Visualization of the evoked potentials averaged over all subjects for both the faces and scrambled faces conditions of the face-detection task in the VEPCON dataset. Also the topography at 160 ms post-stimulus is indicated, as this is considered the peak of the N170 component in the faces condition.

Figure 7

Illustration of the fusiform area and the difference in the reconstructed activity between both conditions for the N170 component averaged over all subjects. For the subject-specific reconstructions, the source activity was morphed to the template head model before averaging. In the case of the subject-specific head models using FEM-based approaches, activity is found in the left and right fusiform area. When using the template head models, on the other hand, the strongest difference in activity between both conditions is found more occipitally.