Effect of spatial smoothing on regions of interested analysis basing on general linear model

Functional magnetic resonance imaging (fMRI) is a powerful, non-invasive method used by psychologists, psychiatrists and neurologists to study the functioning of the brain. The approach of spatial smoothing using a gaussian filter kernel with FWHM is commonly used in fMRI studies. This step is an important as well as a controversial operation with both advantages and disadvantages. For the general linear model analysis, the effect of spatial smoothing on BOLD response amplitude is remain no well know. We will investigate that spatial smoothing effected on response amplitude in the region of LO, which was thought as areas of object-selective response. Using common block design and images of faces, houses, objects and textures, we compared the neural response to these images. We considered satieties map and t value and beta value in a certain regional, with spatial smoothing with a gaussian filter kernel with FWHM sizes 0 to 20 mm with a step of 4 mm. We found the face, house, object exhibited stronger satieties map neural response than texture. In a certain regions, the spatial smoothing effect the t values greatly, smoothing result significant bigger t value than with no smoothing. There were exhibited a bit difference among each category. The spatial smoothing effect the beta values for a certain stimuli, For the face stimuli, smoothing with FWHM>8 result significant smaller beta value than with no smoothing. But the texture had bigger beta values when the spatial smoothing bigger than 4mm. Spatial smoothing had no influence on beta value for house and object stimuli. In conclusion, spatial smoothing is beneficial to the efficacy of analyses, spatial smoothing will affect the beta values for a certain task or certain regions. For our experiment, we propose that the optimal FWHM is about 8mm.