Abnormal Fusiform Gyrus Response to Low but Not High Spatial Frequency Face Information in Autism

Background: Face recognition deficits and hypoactivation in classically face-selective fusiform cortex are widely replicated in ASD. However, little is known about the mechanisms underlying such deficits and whether they might be tied to specific abnormalities in the perceptual processing of face information. For example, electrophysiological studies suggest that there may be abnormal specialization of spatial frequency (SF) processing in the ASD visual system. Notably, low spatial frequency (LSF) information is thought to better support the holistic processing strategies recruited for face recognition among healthy populations, whereas high spatial frequency (HSF) information better supports more featural processing strategies not unlike those recruited for face processing in ASD.

Objectives: To test the prediction that fusiform hypoactivation in ASD stems from an abnormal response to the LSF, but not HSF, components within face stimuli.

Methods: Whole brain fMRI images were collected from 31 participants (16 ASD; 15 age-/IQ-matched controls) while they performed identity-matching judgments on simultaneously presented image pairs. To assess whether the ASD sample showed fusiform hypoactivation for unfiltered faces relative to non-face stimuli, participants completed a standard localizer run with unfilltered face and house images. To examine the neural response to different SF information, participants performed five runs of matching trials with face or shoe images that were either broadpass (BSF; i.e. containing the full spectrum of SF information) or filtered to contain only LSF (<8 cycles/image) or HSF (>32 cycles/image) information.

Analyses were performed on a priori ROIs functionally defined by contrasting BSF faces with BSF shoes for the data collapsed over group. The TAL coordinates for the peak voxels were as follows: right fusiform gyrus (RFG) = (39,-37,-17); left fusiform gyrus (LFG) = (-39,-37,-20). Volume-of-interest analyses were performed on the 100 most face-selective voxels in each region.

Results: As found previously, the ASD group, relative to the control group, showed right fusiform hypoactivation for unfiltered faces (compared to houses) in a location consistent with previous studies (uncorrected p=.005; TAL co-ordinates: 39,-52,-17).

There was a main effect of category in the RFG, and most notably a significant three-way interaction between category, SF scale, and group, F(1,29)=4.70, p=.039: the ASD group showed a similar advantage for HSF faces over HSF shoes as did the control group, but a reduced advantage for LSF faces over LSF shoes. No other effects reached significance.

There was also a main effect of category in the LFG, F(1,29)=13.92 p=.001, with greater activation for faces compared to shoes and an interaction between category and SF scale, F(1,29)=6.49, p=.016, with greater face-selectivity for HSF compared to LSF faces. No other effects reached significance.

Conclusions: Consistent with previous findings of hypoactivation to face stimuli in ASD, face-selective right fusiform activation for LSF images was reduced in ASD relative to healthy controls.  However, face-selective activation for HSF images was comparable among the two groups. The ASD group’s abnormal pattern of face-selective activity to SF filtered face images suggests that the fusiform gyrus in ASD may be abnormally tuned (relative to controls) to the perceptual information in faces.