An Approach To Validation of Diffusion MRI-based White Matter Tractography

A. G. Sorensen, R. Wang, T. Benner, N. Makris Athinoula A. Martinos Center for Biomedical Imaging, Charlestown, MA, United States, Center for Morphometric Analysis, MGH, Charlestown, MA, United States Purpose: The validation of diffusion MRI-based white matter tractography in the human brain is a major challenge. No comparable non-invasive technique exists for cross-validation, and diffusion coefficients change almost immediately after death, making ex-vivo validation problematic. We sought to overcome this by devising a knowledge-based macroscopic validation strategy. This approach draws on the existing knowledge of neuroanatomical connectivity in humans to build a matrix of over 1500 regions of expected connectivity between various brain areas as designated by an expert neuroanatomist based on classical literature-based descriptions of white matter connectivity. This matrix can then be used as a “truth table” to examine various approaches to white matter connectivity. We patterned our approach after a publicly available database of macaque brain connectivity (www.cocomac.org) that in a similar way lists what is known about connectivity of the primate brain. We tested four different streamline-based connectivity approaches to determine their overall accuracy compared to the “truth table”.