Abstract
Cerebral vessel segmentation is essential and helpful for the clinical diagnosis and the related research. However, automatic segmentation of brain vessels remains challenging because of the variable vessel shape and high complex of vessel geometry. This study proposes a new active contour model (ACM) implemented by the level-set method for segmenting vessels from TOF-MRA data. The energy function of the new model, combining both region intensity and boundary information, is composed of two region terms, one boundary term and one penalty term. The global threshold representing the lower gray boundary of the target object by maximum intensity projection (MIP) is defined in the first-region term, and it is used to guide the segmentation of the thick vessels. In the second term, a dynamic intensity threshold is employed to extract the tiny vessels. The boundary term is used to drive the contours to evolve towards the boundaries with high gradients. The penalty term is used to avoid reinitialization of the level-set function. Experimental results on 10 clinical brain data sets demonstrate that our method is not only able to achieve better Dice Similarity Coefficient than the global threshold based method and localized hybrid level-set method but also able to extract whole cerebral vessel trees, including the thin vessels.
Highlights
Cerebral vascular diseases have become the main incentives to dizziness, disability, and even death in many countries around the world, and the research for vessels arouses concern
We propose an active contour model (ACM) implemented by the level-set method in order to segment cerebral vascular structures from TOF-MRA data
The role of the thirdboundary term is equivalent to the geodesic active contour model, and it encourages the contour curve to enclose the regions with high image gradient
Summary
Cerebral vascular diseases have become the main incentives to dizziness, disability, and even death in many countries around the world, and the research for vessels arouses concern. The segmentation of cerebral vascular structures is important for the clinical diagnosis and analysis. In medical image processing field, segmentation means the extraction of anatomical structures of interest from original data [1, 2]. Because of low contrast of images, edge blur, and structure complexity of cerebral vessels, the accurate segmentation is still a challenging task and deserves to be researched [3, 4]. A comprehensive review can be referred to in Lesage et al work [17]. Among these techniques, the ACM has been widely applied in medical image segmentation because of its easy extensibility
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
