Automatic tractometry methods to evaluate brain fiber tracts in dementia

DW-MRI tractography provides a non-invasive, in vivo approach to visualize the brain’s anatomical connections. By characterizing the directional motion of water within brain tissue, it allows inference of the orientation and organization of WM fibers. Tractometry, or along-tract analysis, is a method for studying the microstructural properties of WM pathways by evaluating diffusion MRI metrics along the trajectory of the reconstructed bundles. This approach captures spatial variations along the tract, providing a more detailed characterization of structural integrity and allowing comparisons across subjects, populations, or tractography methods. This work compares two tractography and tractometry pipelines: a custom developed pipeline based on free software tools (MRtrix3 based pipeline) and an automatic pipeline (based on TractSeg). Both were used to reconstruct and compute along-tract profiles of the Arcuate Fasciculus (AF) and Corticospinal Tract (CST) for metrics derived from Diffusion Tensor Imaging (DTI) and Neurite Orientation Dispersion and Density Imaging (NODDI) in 22 subjects selected from the ADNI repository. Significant differences in both tract reconstruction and along-tract profiles were observed across all tracts between the two pipelines. These differences were statistically supported by a cluster-corrected paired permutation t-test comparing the two methods node-wise for each tract. Following this validation, generalized additive mixed models (GAMMs) were fitted separately for each pipeline and tract to model the along-tract profiles and assess the effects of subject group, hemisphere, and tract position on the diffusion metrics. Consistent with previous studies, more extensive significant differences were observed in the AF, while the CST showed limited alterations between AD and CN. The MRtrix3 based pipeline appeared more sensitive in detecting group differences between AD and CN subjects, especially for FA, MD, RD, and NDI.