PROCEEDINGS ARTICLE | March 14, 2018
Proc. SPIE. 10493, Dynamics and Fluctuations in Biomedical Photonics XV
KEYWORDS: Statistical analysis, Molecular bridges, Tissues, Optical coherence tomography, Particles, Bridges, Angiography, Cerebral cortex, Natural surfaces, Alzheimer's disease
Penetrating vessels bridge the mesh of communicating vessels on the surface of cortex with the subsurface microvascular bed that feeds the underlying neural tissue. The degeneration and dysfunction of penetrating vessels directly relates to Alzheimer’s disease, perceptual deficit, amnestic syndrome and stroke. Here we propose a cerebral penetrating vessel mapping approach based on eigen decompensation (ED) principle component analysis that is innovatively redesigned from optical coherence tomography (OCT) angiography. Ensemble complex OCT signals acquired through repeated A-scans first form a covariance matrix and then project into an eigenspace to represent frequency components of moving particles. The eigen representation of signals possesses several advantages over that in spatiotemporal domain: 1) the eigen components possess distinct statistical distributions for penetrating vessels, surface communicating vessels, vessel free regions, and territories occupied by enriched capillaries; 2) this approach is immune to tailing artifacts, enabling automatic decoupling of penetrating vessels from lateral vasculature networks. To describe the uniqueness of penetrating vessels as 2D parameter mapping, a second round of eigen analysis is applied to the eigen representations by taking each eigen component as an observation and distributions of the eigen components as features. In our datasets of mouse cerebral cortex, the eigen components mainly follow a subtle logistic distribution, statistically more significant than other features in terms of distribution spectral power (> 30 dB). While, the existence of vessel penetrating behavior locally breaks this distribution, assigning low transform probabilities to corresponding A-scans. Therefore, the transform coefficients inversely correlate to the vessel penetration and fully reveal the spatial morphology of penetrating vessels from projection view. This method allows for automatic statistical quantification of penetrating arterioles and ascending venules from large volume OCT angiography data, and accordingly contributes to the morphometric analysis of cortical microvasculature in functioning brains.
