Mapping and Quantitating Penetrating Vessels in Cortical Brain Using Eigen-Decomposition of OCT Signals and Subsequent Principal Component Analysis

Penetrating vessels bridge the mesh of communicating vessels on the surface of the cortex with the subsurface microvascular beds that feed the underlying neural tissue. Their accurate identification in vivo is important in the investigations of neural degenerative diseases, e.g., Alzheimer's disease and stroke. Here, we propose an efficient method to automatically map cortical penetrating vessels based on an eigen decompensation analysis of the optical coherence tomography (OCT) and OCT angiographic signals. We first project the ensemble of repeated OCT signals into a feature space that represents the power spectral components of eigenvectors through a well-known eigen-decomposition method. A principal component analysis is then applied to the spectral components to identify penetrating vessels. We find that the spectral components of OCT signals captured from a cortical brain tissue follow a subtle logistic distribution, which is, however, broken down if there are penetrating vessels. Such feature allows for an automatic mapping of penetrating arterioles and ascending venules from large volume OCT-scan datasets, accordingly contributing to the topological and morphological analyses of cortical microvasculature in the functioning brains. To demonstrate the utility of the proposed method, an OCT angiography imaging platform is used to show the functional behavior of penetrating blood flows before and after an ischemic insult in an established middle cerebral artery occlusion (MCAO) model of rodent, where an average of 41% reduction in penetrating vessel density (n = 5 animals) was observed in the ischemic region post-MCAO.