Introduction Recent studies show an atypical perceptive behavior on face processing in autism. Functional studies show that the fusiform gyrus is involved in face perception and is not or less activated in autistic subjects [1]. These face processing is believed to affect social interaction, which can be quantified by ADOS scores and related to the functionality of the prefrontal cortex. The aim of this study is to quantify anatomical changes in autistic subjects on the white matter tracts that traverse the fusiform gyrus, the prefrontal cortex and the superior temporal gyrus and correlate them with ADOS scores to complete. In this study, we first verified the connectivity difference between the fusiform gyrus, the prefrontal cortex and the superiotemporal gyrus through fMRI. Then, we used Diffusion Tensor MRI (DT) images to assess the integrity of automatically segmented white matter bundles connecting these girii. Finally, we perform statistical analysis on these fiber bundles using diffusivity measures calculated from DT to characterize tissue microstructure changes and correlate these changes with ADOS scores. Materials and methods Data acquisition. We acquired dMRI of eight adults with high-functioning autism or Asperger syndrome and eleven healthy adults at the University of Buenos Aires on a GE Signa Hdxt 3.0T scanner. The acquisition consisted in 80 directions with b=1000s/mm2 and 1 image with b=0s/mm2 with 1×1×1.3mm voxel size. Data preparation. We started by generating an unbiased template of the DT images and registering linearly and non-linearly all of the images to it. Then, we computed DT-based full brain tractography for every subject. Finally, we used the tools developed by Wassermann et al [2] to cluster fiber bundles and select the bundles traversing the fusiform gyrus. This gyrus was identified through the girii parcelation of the JHU atlas. We clustered these bundles across subjects extracting a set of population-obtained bundles (see for Fig 1 for an example). Data analysis. We applied the statistical analysis of Wassermann et al [3] for each population-obtained bundle, we calculated its tract-probability map (TPM) (see Fig.3 for an example) and skeletonized this map to obtain a bidimensional representation of each bundle. For each patient, we projected the measure of diffusivity (FA, axial or radial diffusivity) around the tract to their closest point on the skeleton and we average them with a weight according to the TPM. This produces two populations (one for autists and one for controls) of projected functions on the skeleton. For each tract and each measure, we used a cluster-based permutation hypothesis testing approach [4] to detect dissimilarities between both populations. For significant clusters, we calculated correlation between the mean diffusivity measure and ADOS scores. Results We found different clusters (Fig 1, red voxels) where there were dissimilarities between autistic and healthy subjects in FA measures on tracts traversing the fusiform gyrus in both hemispheres of the brain. We observed a reduction of FA values in a cluster (Fig 2, red voxels) on a bundle joined the superior temporal gyrus to the prefrontal cortex. In this cluster, the mean FA on the autistic subjects was strongly correlated with their scores to ADOS social interaction with a correlation coefficient lower than -0.96 a p-value lower than 0.004 (Fig3). Discussion First, results revealing difference between controls and autists in clusters on fiber bundle traversing the fusiform gyrus are in agreement with current literature giving this area as a classical area of face perception. Agreement reinforced by functional study of autists showing a drop of connectivity between the fusiform gyrus and the superior temporal cortex in comparison with controls (Fig 4). The localisation of the cluster where we founded a correlation with ADOS social interaction scores (Fig3) is in agreement with current anatomical literature, the superior temporal gyrus being a classical area of face perception and emotion and the prefrontal cortex a classical area of social interaction. This is reinforced by the strong inverse correlation showing that in this linking cluster a drop of FA value is strongly correlated by a rise of the intensity of autism. Our fMRI study (Fig 4) corroborates these statistical results, showing a drop of connectivity between the prefrontal cortex and the superior temporal gyrus. Disruption of white matter tracts between regions implicated in social functioning and face perception may contribute to increase the severity of autism.