A geometric flow for segmenting vasculature in proton-density weighted MRI

Abstract

Modern neurosurgery takes advantage of magnetic resonance images (MRI) of a patient's cerebral anatomy and vasculature for planning before surgery and guidance during the procedure. Dual echo acquisitions are often performed that yield proton density (PD) and T2-weighted images to evaluate edema near a tumor or lesion. In this paper we develop a novel geometric flow for segmenting vasculature in PD images, which can also be applied to the easier cases of MR angiography data or Gadolinium enhanced MRI. Obtaining vasculature from PD data is of clinical interest since the acquisition of such images is widespread, the scanning process is non-invasive, and the availability of vessel segmentation methods could obviate the need for an additional angiographic or contrast-based sequence during preoperative imaging. The key idea is to first apply Frangi's vesselness measure [1] to find putative centerlines of tubular structures along with their estimated radii. This measure is then distributed to create a vector field which allows the flux maximizing flow algorithm of [2] to be applied to recover vessel boundaries. We carry out a qualitative validation of the approach on PD, MR angiography and Gadolinium enhanced MRI volumes and suggest a new way to visualize the segmentations in 2D with masked projections. We validate the approach quantitatively on a single-subject data set consisting of PD, phase contrast (PC) angiography and time of flight (TOF) angiography volumes, with an expert segmented version of the TOF volume viewed as the ground truth. We then validate the approach quantitatively on 19 PD data sets from a new digital brain phantom, with semi-automatically obtained labels from the corresponding angiography volumes viewed as ground truth. A significant finding is that both for the single-subject and multi-subject studies, 90% or more of the vasculature in the ground truth segmentation is recovered from the automatic segmentation of the other volumes.

Angiographic Image Analysis

“…One of the most popular hybrid methods is combination of multiscale differential analysis within vessel detection schemes as in [34,84] with deformable models, such as levelsets [15], Bspline snakes [33] and maximum geometric flow [24,103].…”

Filtering and segmentation of 3D angiographic data: Advances based on mathematical morphology

“…Non-exhaustively, one can cite: region-growing [74], deformable models [41,15], statistical analysis [12,62], minimal path-finding [37], vessel tracking [22,43], differential analysis [66], or mathematical morphology (discussed in Section 2.1). Despite this wide range of methodological contributions, the results provided by segmentation methods generally remain perfectible.…”

Segmentation of Intracranial Vessels and Aneurysms in Phase Contrast Magnetic Resonance Angiography Using Multirange Filters and Local Variances

“…This descriptor can be regarded as the computation of inward flux of the image gradient [16], [17]. It is widely employed on various vascular detection frameworks [16]- [21].…”