The corpus callosum (CC) has long been investigated in autism spectrum disorder (ASD)¹, suggesting a hypothetically important role of aberrant interhemispheric connectivity in this mental disorder. Previous neuroimaging studies on brain structures have consistently reported reductions of the tract integrity in ASD as compared to typically developing (TD) subjects in the period of adolescence and adulthood. Few have investigated the developmental difference of the CC between ASD and TD. In this study, we performed diffusion spectrum imaging (DSI) to estimate the generalized fractional anisotropy (GFA) value which was considered to reflect the tract integrity. A template-based tract-specific analysis named tract-based automatic analysis (TBAA)² was used to explore structural development of the CC in ASD participants over a wide range of age. We hypothesized that there was age-related abnormalities of the CC in ASD.Eighty-five males with ASD and 96 age and handiness matched TD participants (ASD: 15.02 ± 4.37, range: 7.8-27.8 years old; TD: 15.82 ± 5.66, range: 8.8-29.0 years old), were recruited in the study. Images were acquired on a 3T MRI system with a 32-channel head coil (Tim Trio, Siemens, Germany). DSI was performed using a twice-refocused balanced echo diffusion echo planar imaging sequence (TR/TE = 9600/130 ms, image matrix size = 80 x 80, spatial resolution = 2.5 x 2.5 mm^2, slice thickness = 2.5 mm, and 102 diffusion encoding gradients with maximum bmax = 4000 s/mm^2). For DSI analysis, TBAA method was applied via a high quality DSI template and predetermined white matter tracts in the whole brain. The DSI template was constructed by coregistering 122 healthy participants’ DSI datasets using a registration method under the Large Deformation Diffeomorphic Metric Mapping (LDDMM) framework³. Sixteen white matter tracts of the CC were reconstructed on the template using multiple regions of interest defined in the Automatic Anatomical Labeling system. Each reconstructed tract was subdivided into 100 steps with even spacing and the step coordinates along tracts were saved as sampling coordinates. The procedures of the TBAA method were as follow. 1) A study specific template (SST) was created via coregistering all participants’ DSI datasets using LDDMM. 2) The SST was coregistered to the DSI template. 3) Sampling coordinates of the tracts were transformed from the DSI template to individual DSI datasets via the transformation matrix between DSI template and SST as well as the matrix between SST and individual DSI. 4) GFA values were sampled in the native DSI space using the transformed sampling coordinates, resulting in a 2D array (16 tracts x 100 steps) of GFA profiles for each subject. In this study, mean GFA for each segment of the CC was chosen for our purpose. To minimize multiple testings, 16 tracts were classified into 7 portions according to their cortical connections, from frontal lobe - prefrontal part, frontal lobe - motor part, paracentral gyrus, parietal lobe, temporal lobe, limbic lobe, to occipital lobe. A hyperbolic model was applied to fit the age-GFA nonlinear model for each portion. F test was applied to test the difference in model comparison between ASD and TD. Benjamini-Hochberg testing was applied to correct multiple comparisons.Among the 7 portions of the CC, only the temporal lobe portion showed significant difference in age effect between ASD and TD (p = 0.049, corrected) (Figure 1). This portion comprised three commissural tracts, i.e. the CC to bilateral superior temporal lobes, the CC to bilateral middle temporal lobes, and the CC to bilateral temporal poles (Figure 2). The fitted age-GFA curves showed delayed growth pattern in ASD from the period of adolescence to early adulthood.In this study, we performed a tract-specific analysis of the callosal fibers in ASD and TD participants over a wide range of age to investigate the growth pattern of the interhemispheric connections. The CC connecting bilateral temporal lobes showed significant difference in the age-GFA model between the two groups. Delayed growths of the GFA values were found in the period of adolescence and early adulthood in ASD group. As GFA values are often considered to reflect microstructural integrity of the white matter, the finding may suggest a maturation delay in the CC to bilateral temporal lobes in patients with ASD. The maturation delay may implicate the deficits regarding to the functions of the temporal lobes, such as language, semantics, and sensory processing in ASD patients.An atypical age-GFA pattern found in the CC connecting bilateral temporal lobes suggests a developmental delay of the white matter tracts in the CC in ASD patients. A longitudinal study is warranted for further investigation.