Optimal Face Network Localization in Autism: A Comparison of Two Methodologies

Background:  

The vast majority of studies looking at category-specific processing identify regions of interest (ROIs) based on univariate localizer contrasts. Information regarding optimal localizer choice for a particular purpose (such as connectivity or multi-voxel pattern analysis) or population of interest (such as autism) is necessary for efficient data acquisition and optimal analysis implementation. Because hypoactivation in the face processing network has been independently replicated in autism, localization of these regions can be problematic. We compare two extremely different localizers that differ in a variety of ways and evaluate their capacity to activate regions in the face-processing network of subjects with an autism spectrum disorder (ASD).

Objectives:  

To determine and compare the effectiveness of two localizing methodologies for regions involved in face processing in an ASD population.

Methods:  

Whole brain fMRI images were collected from 10 male ASD participants (mean age: 14; mean IQ: 104) while they performed two tasks.  The first task consisted of rapidly presented faces, places, objects, and scrambled objects.  Face images consisted of uncropped, colorful, visually distinct, attractive faces of a variety of ages and ethnicities while places were colorful, visually distinct, scenes from unidentifiable locations.  Images were not controlled on any low-level visual dimensions and were presented in 14-second blocks with 18 images per category per block. Participants performed an orthogonal task, which was to respond to a white fixation cross that appeared in the center of an image. In the second task, participants performed identity-matching judgments on simultaneously presented face or house image pairs.  Images were well-controlled, black-and-white pictures of cropped faces and houses, presented at a rate of 10 images per 20-second block.  Random effects analysis of activation differences between tasks was performed on whole-brain images.  In order to additionally investigate differences between these two methodologies, we quantified a series of metrics on regions of the face-processing network (including fusiform face area (FFA), occipital face area (OFA), superior temporal sulcus (STS), and amygdala).  We present data on both whole-brain and region of interest analyses including peak activation, mean activation, voxel extent, and voxel overlap between participants on each localizer task.

Results:  

Whole brain analyses on the rapid presentation localizer revealed significant activation in ventral visual cortex (p<.05 FWE corrected).  Results of the discrimination localizer did not reach whole brain significance.  However, both tasks were successful in region of interest analyses using individually defined clusters of activation.

Conclusions:  

Our findings suggest that a rapid-presentation, less visually controlled (colorful images of varying perspectives) paradigm with an orthogonal task is effective for localizing visual cortical nodes of the face processing network.   Significance at the whole brain level in the rapid presentation localizer suggests this may activate regions of ventral visual cortex that are more likely to overlap across individuals. This may be due to the increased number of images presented in this task or the engaging nature of the stimuli used.  Our results are informative for making localizer choice decisions, particularly in regard to advanced analyses such as connectivity and multi-voxel pattern analysis.

See more of: Brain Imaging-Functional
See more of: Brain Imaging: fMRI-Social Cognition and Emotion Perception
See more of: Brain Structure & Function

| More