Meeting Banner
Abstract #0703

Fiber Density Estimation from Single Q-Shell DWI by Tensor Divergence

Marco Reisert1, Elias Kellner1, Henrik Skibbe1, Valerij Kiselev1

1Medical Physics, University Medical Center, Freiburg, Baden-Wrttemberg, Germany

Diffusion-sensitized magnetic resonance imaging provides information about the nerve fiber bundle geometry of the human brain. While the inference of the underlying fiber bundle orientation only requires single q-shell measurements, the absolute determination of their volume fractions is much more challenging. This work uses a conservation law for fiber orientation densities that can infer the absolute fraction. In this study we show by simulations on a pseudo ground truth phantom that for complex, brain-like geometries the method is able to infer the densities correctly. In-vivo results with 81 healthy volunteers scanned with a clinical feasible protocol are plausible and consistent.

Abstract PDF Presentation Video

Keywords

able absolute accessible account achieve although anatomical angular appear arbitrary assumed assumes assumption baden boundary brain bundle bundles certain challenging characteristics classical clinical complex conditional conditions conservation consider consistent constant continuity coordinate cope corpus correct cost crossing curvature densities density derived derives detailed determination diffusion directly discretization distribution divergence dynamics element equals equation estimation even experiment feasible fiber fibers field find finite fluid fraction fractions function functional generated geometries geometry global goal gradient ground hand healthy human idea in vivo indicators individuals infer inference inferred inspired integral integrated joint knowledge leave like locally magnitude major mass medical minimized minutes much nerve observable optimization orientation orientations particular percentage phantom physics preservation principle proceedings projection proposed protocol pseudo quantity randomly rather related require requires respect scanned secondly selected sensitized shell simulations slices solved source spinal straight subjects suggest superior taken tensor terminate terms tracts transversal truth undergoes underlying unknown untouched validity vector volume volunteers white