Co-activation patterns in somatosensory areas: A meta-analysis study

Abstract

Event Abstract Back to Event Co-activation patterns in somatosensory areas: A meta-analysis study Lars Forsberg1* and Per Roland1 1 Karolinska Institute, Sweden Introduction: NeuroGenerator is a project that generates statistical image databases from PET and fMRI studies (Roland et al. 2001). This is done by processing the raw data in a uniform manner. The statistical database contains statistical Z-images and cytoarchitectural population maps in Talairach space. In this study, we use the statistical images to find co-activation patterns for experiments engaging the somatosensory area 3b. Methods: We first made a voxel-wise correlation between all the statistical Z-images and the population map of cytoarchitectural area 3b. The statistical image with the highest correlation had a correlation of 0.20. This image was then used to calculate the voxel-wise correlation with all the other statistical images in the database. Nine statistical images were found to have a higher correlation than 0.10, which was considered fairly high given the small ratio between the size of the activations and the total size of the brain. A sum image was calculated from these images and then normalized in order to regard the noise as ~N(0,1). Finally, we used False Discovery Rate to find the threshold value for a corrected p = 0.01 in the normalized sum image. Results: The threshold level for the normalized sum image was found to be Z ≥ 2.90, which corresponds to an uncorrected p = 0.002 and a corrected p = 0.01. The results show significant activations in the left hemisphere for areas 3b, 4a and 4p (figure 1), and significant bilateral activations for the intraparietal areas ip1 and ip2 (figure 2). No activation was found in area 6. Conclusion: The experiments in the study have in common that they engage area 3b in the left hemisphere. For these experiments, we also found a bilateral co-activation pattern for the areas ip1 and ip2. The chosen method is sensitive to activations that are weak and unsignificant in each individual experiment but reproduced across experiments. There is a clear benefit of such a method compared to meta-analyses of databases with published functional neuroimaging experiments, where only reported findings from each study can be used in the meta-analysis.