Random forests for dura mater microvasculature segmentation using epifluorescence images

Abstract

Automatic segmentation of microvascular structures is a critical step in quantitatively characterizing vessel remodeling and other physiological changes in the dura mater or other tissues. We developed a supervised random forest (RF) classifier for segmenting thin vessel structures using multiscale features based on Hessian, oriented second derivatives, Laplacian of Gaussian and line features. The latter multiscale line detector feature helps in detecting and connecting faint vessel structures that would otherwise be missed. Experimental results on epifluorescence imagery show that the RF approach produces foreground vessel regions that are almost 20 and 25 percent better than Niblack and Otsu threshold-based segmentations respectively.

Estrogen-Dependent Changes in Dura Mater Microvasculature Add New Insights to the Pathogenesis of Headache

“…To analyze vascular integrity, a perfused dura mater model was observed using high-resolution digital epifluorescence, laser scanning confocal microscopy and computer-assisted morphometric analysis ( 31 ). The results of the advanced microscopic analysis revealed that large areas of the connective tissue stroma, thought previously to be avascular, possessed extensive microvascular networks ( 32 – 37 ). Further, the cessation of the ovarian hormone production following bilateral ovariectomy in Yucatan miniature pigs caused gradual (within 2 months) yet significant remodeling of these microvascular networks.…”

Confocal vessel structure segmentation with optimized feature bank and random forests

“…We use random forest (RF) classifier [8] based approach which is adapted for vessel structure and cell segmentation with high precision and recall [7], [9]. Our feature bank is optimized to capture curvilinear structures of varying width as well robust to scale changes.…”

“…They are based on Hessian eigenvalues based multiscale Frobenius norms, multiscale Laplacian of Gaussians (LoG), a multiscale line detector [2]. Here we provide the basic details of each of these features and refer to [7] more details of these optimized feature bank for vessel segmentation.…”

“…To avoid these drawbacks, in [7] we used a set of vessel specific optimized features along with random forest (RF) classifier based segmentation for robust curvilinear structure extraction from epifluorescence imagery under challenging imaging conditions. In this work, we extend our work in [7] by utilizing intensity based Laplacian of Gaussian (LoG) for capturing the spatial variations which are unique to confocal imagery.…”

“…In this work, we extend our work in [7] by utilizing intensity based Laplacian of Gaussian (LoG) for capturing the spatial variations which are unique to confocal imagery. Due to the nature of confocal imaging where local intensity information is important, adding a multiscale LoG based intensity feature improves the final segmentation accuracy.…”