Reconstruction of Three-Dimensional Nerve Fiber Orientations from Histological Brain Sections in Three-Dimensional Polarized Light Imaging

Abstract

Three-Dimensional Polarized Light Imaging enables to reconstruct nerve fiber orientations from polarimetric measurements of histological brain sections based on the birefringence of myelinated nerve fibers. Measurements of the brain sample from oblique views facilitate the retrieval of the three-dimensional orientation of the nerve fibers as well as the birefringence strength. The enhancement of this noise sensitive reconstruction represents the major motivation for this thesis. First, a novel and fast reconstruction algorithm based on a least squares approach is introduced. At the mesoscale an unprecedented comprehensive view of the human brain’s nerve fiber tracts is obtained based on the new reconstruction algorithm. Next, the reconstruction is further improved for very low signals utilizing a Bayesian estimator. Furthermore, as the parameter estimation is sensitive to noise, the uncertainty of the obtained parameter maps is studied. This results in the first nerve fiber orientation confidence estimates for Three-Dimensional Polarized Light Imaging. Finally, the developed analysis is applied at the microscale. A validation against higher-resolved volumetric measurements shows that individual fiber bundles can be reconstructed with high accuracy using the developed algorithms.