Meeting Banner
Abstract #3151

Assessment of Local Fiber Plausibility Using a HARDI Based Reference

Peter F. Neher1, Bram Stieltjes2, Ivo Wolf3, Hans-Peter Meinzer1, Klaus Hermann Fritzsche1, 2

1Medical and Biological Informatics, German Cancer Research Center (DKFZ), Heidelberg, Germany; 2Quantitative Image-based Disease Characterization, German Cancer Research Center (DKFZ), Heidelberg, Germany; 3Medical Informatics, Mannheim University, Mannheim, Germany

Evaluation of diffusion MRI-based tractography methods is challenging and existing approaches for quantitative assessment mainly focus on measures of general tract overlap or the presence of anatomically plausible connections between brain regions. We present an evaluation of the local fiber direction plausibility on a voxel-wise basis. We further propose a reference-based validation procedure that aims at evaluating tractograms obtained from data of clinical quality using high quality HARDI data as a reference.

Keywords

acquisition additionally allow allows angular applications applied applying approaches artificial artificially assessment automated automatic ball benefit bias biological brain bundles cancer challenging characterization chose clearly clinical comparably configuration configurations connections consistency constrained constructed containing coronal correctly course crossing datasets defined depicting describe deterministic development diffusion directly disease diversity enables enhance error evaluated evaluates evaluation even example expected extracting fact fashion fiber fibers framework furthermore future general generated global gold gradient healthy highest human important in vivo informatics insight integration like limitation limited living local mainly manually many measure measures medical medicine missing modeling often open optimized original owed performance peter phantom planning plausibility pons probabilistic procedure proposed prototype pushing quality quantitative ranging reconstruction reduction reflected regarding relatively represent represented repulsion resembles restricted retrieved reveals segmentation sets shell simple since slice source spherically streamline streamlining strengths structured subjects surface systems tasks tend tensor towards tracked tract underlying validation weaknesses whether white wise wolf yielding