We focused on analysis of long-term medication influenced white matter tracts based on diffusion-weighted MRI brain images of patients with ADHD disorder. We applied a framework of consistent model selection with Tract-based spatial statistics (TBSS) to give proper and consistent modelling of fiber-crossing in white matter. An orientation atlas was constructed to give an `orientation prior' during the ball-and-2sticks model estimation. Consistent metrics of fiber properties were obtained for each subject and thus statistical power in crossing-fiber region was improved. Besides, we improved this pipeline by optimising orientation prior and by integrating a high-dimensional image registration.