Diffusion Tensor Tracking (DTT) of Face Processing Pathways with Comparison to Sensitive Behavioral Measures: Understanding Microstructure and Detecting Subgroups in Autism

Background: Children and adults with autism have characteristic impairments in recognizing faces and face emotions.  Diffusion tensor tracking (DTT) enables detection/measurement of previously-unknown hippocampo-fusiform (HF) and amygdalo-fusiform (AF) pathways  (Smith et.al., 2009), respectively involved in face recognition and emotional processing.  Given the importance of face processing in autism, we analyzed HF/AF pathways in adolescents/young adults with autism (Conturo et al., 2008).  The right HF pathway (specific for face recognition) had reduced across-fiber diffusivity (D-min).  To better understand the microstructural basis for these changes, we compared DTT to sensitive neuropsychological tests (NPTs) of face recognition/emotional processing designed for study of autism.  By characterizing participants on two dimensions (2D), participants might also separate into subgroups.

Objectives: Interpret DTT abnormalities and test for subgroups by comparing to sensitive NPTs.

Methods: Custom diffusion-tensor MRI data were acquired in 17 participants with high-functioning autism meeting ADOS/ADI criteria (age 16-53) and 17 controls (individually-matched in age/gender/ethnicity/handedness, and verbal/performance/full-scale IQ).  Group differences in right HF D-min (p=0.014, paired t-test) were analyzed by administering to 10 of the same autism participants sensitive NPTs of face-gender identification (Wilkinson et.al., IMFAR2009) and face-emotion recognition (Rump et.al., 2009).  Stimulus faces were digitally morphed together to produce subtle gender or emotion (e.g., 40%/60% male/female).

Results: Both the face-gender and face-emotion NPTs showed a strong relation to DTT for both right HF/AF, with slower D-min related to lower performance in all cases.  The correlation between D-min and gender identification accuracy was modest (r=0.42/0.42 for right HF/AF), with similar results for emotion.  A 2D separation of participants into two subgroups was revealed: one subgroup (5 subjects) had low performance with a very low pathway D-min; the other subgroup had higher performance with a slightly higher D-min.  Both subgroups exhibited similar DTT-NPT slopes, with very high correlation in the low-performance subgroup (0.93/0.87 for right HF/AF). [The high-performance subgroup had lower correlation (0.33/0.57).]  The consistent association of slower D-min with lower performance indicates that the reduction in right HF D-min in (Conturo et.al., 2008) is functionally significant, with the same process likely occurring in right AF.  This result is intriguing, as decreased D-min (and a corresponding anisotropy increase) is typically related to higher performance (i.e., more myelination).  These results reinforce the microstructural mechanism that we proposed in (Conturo et al, 2008): small-diameter axons in the affected pathways, causing slowed across-fiber diffusivity and slowed neural transmission.  Small diameters could parsimoniously account for the unusual combination of reduced D-min and reduced function.  The subgroup analysis suggests that some individuals with less severe diameter reduction can develop mechanisms to achieve normal face processing despite their pathway microstructure.  (Such intervening variables likely reduce the correlation in the high-functioning subgroup.)  This interpretation is consistent with the known social histories of the participants.

Conclusions: The strong association between decreased D-min and decreased function in face processing pathways supports the mechanism of small-diameter axons in autism. Sensitive DTT measures of autism-relevant pathways, combined with sensitive NPT measures of pathway function, has potential to elucidate autism mechanisms and reveal autism subtypes.