White matter (WM) tractography with diffusion MRI (dMRI) has been used to infer anatomic connections between cortical regions. Most brain structural connectivity is established by two procedures, cortical parcellation and WM fiber tracing from a certain parcellated cortical region serving as a seed region of interest (ROI) for tractography. Unlike tracing from inside WM, a recent study [1] revealed that roughly 50% of the long-range WM fibers are inaccessible due to dense WM zones immediately beneath the cortex. These dense WM zones impede tracking from the cerebral cortex. There are two practical approaches to address this problem. First, to dilate the cortex into the adjacent WM to initiate tracing; and secondly, to bypass these dense WM zones and trace directly from WM interior to these dense WM zones [2]. In this study, we evaluated for different developmental age groups, how different dilation distances from cortex and inside WM would affect tractography results used to quantify brain connectivity.Subjects and data acquisition: 3 healthy children (2 F and 1 M; 8.7±1.5 yrs), 3 adolescents (3 F; 16.2±0.7 yrs), and 3 young adults (2F and 1 M; 23±1.9 yrs) were scanned on a 3T Philips Achieva system. The diffusion weighted imaging (DWI) parameters were: single-shot EPI with SENSE factor = 2.3, FOV=224/224/143mm, imaging matrix = 112x112, axial slice thickness = 2.2mm without gap, 30 directions, b=1000 s/mm², repetition=2. T1-weighted (MPRAGE) image with FOV=256/256/160mm and resolution 1x1x1mm was also acquired. Cortical parcellation: Based on T1-weighted image of each subject, cerebral cortex of each hemisphere was parcellated into 33 gyri [3] using FreeSurfer (http://surfer.nmr.mgh.harvard.edu). Selection of the cortical seed ROIs to initiate tracing: The parcellated cortical gyri from the T1-weighted image were transformed to dMRI space to serve as the seed ROIs for tractography. Affine transformation was applied to reorient and transform the parcellated cortical labels into dMRI space with Diffeomap (http://www.mristudio.org). For the first evaluation approach, the precentral gyrus and superior frontal gyrus were dilated 1, 2, 3, 4 and 5 mm into the WM and dilated cortical gyri were used as seed ROIs to initiate WM tracing. For the second evaluation approach, cortical gyral regions were first dilated by 2 mm and 4 mm into the WM, then the cortical gyral regions dilated by 1 mm were subtracted to bypass the dense WM zones. Tractography initiated from a cortical seed ROI: Tractography was conducted with the seed ROIs specified above using Probtrackx in FDT of FSL (http://www.fmrib.ox.ac.uk/fsl). A segmented WM mask (SPM FAST, http://www.fil.ion.ucl.ac.uk/spm) was generated to filter out the spurious tracing entering non-WM regions. Evaluation of tractography results with Dice ratios: To quantify the differences of WM traced from different seed ROIs, we calculated Dice coefficients [4] between traced WM regions with the first or second approach and traced WM regions from corresponding parcellated cortical gyrus with a dilation of 5mm. ANOVA was conducted with the Dice ratios among three age groups to test age effects. Dice ratios between the WM tracts traced from the dilated ROI with the 3 mm dilation and those traced with the dilated ROI with 5mm dilation were calculated. Fig. 1 shows that seed ROIs with larger cortical dilation distance and dilation into deeper WM yield more traced WM for the representative subject in all three age groups. As shown in Fig. 2, the Dice ratio for all 3 age groups increased as a function of dilation distance from the cortex. However, after a certain cortical dilation distance of 3mm, a plateau is reached (Fig. 2). Tracing directly from WM causes incomplete WM coverage similar to that tracing from pure parcellated cortical region (Fig. 2). Fig. 3 reveals there is no significant difference in mean Dice ratios among the three age groups (p=0.19) for tractography from two representative cortical ROIs.Changing the cortical seed ROIs led to dramatic differences in tractography results. Tractography directly from the parcellated cortical ROI without dilation may lead to significant bias and incomplete coverage of the WM tracts connected to the cortical ROI. Tractography from WM immediately close to the cortical region results in incomplete tracing outcome too. The same strategy for selection of cortical seed ROIs can be applied to diffusion MRI of all three developmental age groups. Dilation of 3mm from the parcellated cortical ROI seems to be optimal for tractography from the cortical ROIs. Evaluation of the effects of the cortical ROIs on quantification of the connectivity strength (i.e. number of fibers, tracing probability) and connectivity matrices is under way.