Structure tensor analysis of serial optical coherence scanner images for mapping fiber orientations and tractography in the brain

Quantitative investigations of fiber orientation and structural connectivity at microscopic resolution have led to great challenges for current neuroimaging techniques. Here, we present a structure tensor (ST) analysis of ex vivo rat brain images acquired by a multicontrast (MC) serial optical coherence scanner. The ST considers the gradients of images in local neighbors to generate a matrix whose eigen-decomposition can estimate the local features such as the edges, anisotropy, and orientation of tissue constituents. This computational analysis is applied on the conventional- and polarization-based contrasts of optical coherence tomography. The three-dimensional (3-D) fiber orientation maps are computed from the image stacks of sequential scans both at mesoresolution for a global view and at high-resolution for the details. The computational orientation maps demonstrate a good agreement with the optic axis orientation contrast which measures the in-plane fiber orientation. Moreover, tractography is implemented using the directional information extracted from the 3-D ST. The study provides a unique opportunity to leverage MC high-resolution information to map structural connectivity of the whole brain.