Network Model of Face Processing In Autism Based on Diffusion Tensor Tracking (DTT) and Behavioral Data

Background: Recognition of faces/face emotions is commonly impaired in autism.  Using DTT, we identified previously-unknown hippocampo-fusiform (HF) and amygdalo-fusiform (AF) face processing pathways (Smith et.al., 2009, JMRI).  We found reduced minimum-diffusivity (D-min, intrinsic across-fiber diffusivity; Smith et.al., 2009) in right HF and right AF in autism, which correlated with decreased ability to recognize faces and their emotional content on neuropsychological tests (NPTs) (Conturo et.al., 2008, JINS; Conturo et.al., IMFAR2010).  Such D-min reductions and DTT-NPT relationships suggest a new mechanism of autism -- small diameter axons that cause slowed neural transmission.  This biologic mechanism is also consistent with: small cell bodies in hippocampus (Bauman et.al., 2005) and minicolumns (Casanova et.al., 2002); reduced fMRI correlations (Just et.al., 2004; Kleinhans et.al., 2008); lengthened reaction times (Townsend et.al., 1996); slowed electrophysiology (McPartland et.al., 2004); and symptom abatement with fever (Curran et.al., 2007). 

Objectives: To confirm DTT-NPT relationships and devise a network model of slowed pathway transmission causing decreased ability to recognize faces/face emotions.

Methods: DTT data acquired in adolescent/adults with high-functioning autism meeting ADOS/ADI criteria were compared to additional NPT data acquired using sensitive custom tests of face-gender (Wilkinson et.al., IMFAR2009) and face-emotion (Rump et.al., 2009) recognition.  A network model was developed accordingly.

Results: The gender/emotion NPTs showed a strong associated between lower performance and lower pathway D-min for both right HF/AF pathways (e.g., r=0.589; AF vs. face-gender NPT).  These results confirm that the D-min reduction in autism-vs-controls in (Conturo, et al., 2008) is functionally significant, with the unusual reversal of the expected DTT-NPT relation supporting the small-diameter mechanism.  The behavioral effects can be explained by a network model.  Face visual-feature information first enters primary visual cortex (V1), then passes through the ventral-stream processing stages to arrive at fusiform face area. We posit a parallel passage of face information from V1 to amygdala [via the amygdalo-calcarine pathway in (Lori et.al., 2002)] to fusiform (via AF pathway). This parallel path is analogous to the putative "short-cut" of Rudrauf et.al. (2008) where visual information takes a short-cut to fusiform.  In our model, emotional modulation of face perception occurs by visual information passing from V1 to amygdala and then to fusiform, arriving in fusiform at the same time (or earlier) that feature information arrives via ventral stream.  If the AF were slowed by small-diameter axons (as in autism), emotion information would arrive at fusiform too late to modulate the featural information arriving from ventral stream, as that information would have already exited to other brain areas.  A similar path involves hippocampus for memory modulation.

Conclusions: The association between decreased D-min in right HF/AF, and decreased recognition of faces/face emotions in autism, can be explained by small-diameter axons and slowed transmission in right HF/AF, combined with a network model whereby slowed transmission causes memory/emotional information to arrive at fusiform too late to modulate face information arriving via ventral stream.

Funding:  NIH MH090494; Autism Speaks AS03799; The Nancy Lurie Marks Family Foundation; ACE HD055748; NS039538; DC006691.