Hierarchical Segmentation of Thin Structures in Volumetric Medical Images

Abstract

Abstract. We introduce a new method for segmentation of 3D medical data based on geometric variational principles. A minimal variance criterion is coupled with a geometric edge alignment measure and the geodesic active surface model. An efficient numerical scheme is proposed. In order to simultaneously detect a number of different objects in the image, a hierarchal method is presented. Finally, our method is compared with the multi-level set approach for segmentation of medical images.

An automatic variational level set segmentation framework for computer aided dental X-rays analysis in clinical environments

“…Therefore, they are efficient for extracting an object in an image even if the object consists of several disconnected regions. In order to extract multiple objects from the same image, people normally use hierarchical [10,11,14] or coupled [19][20][21][22] level set methods. In [23], Li et al show that the coupled level set method tends to have higher error than the hierarchical level set method even though the coupled level sets functions may take more time to converge.…”

“…We refer the reader to the survey articles [5][6][7] for more details. Variational level set method was first proposed by Zhao et al [8] and is often used for segmentation [9][10][11]. The variational level set method approach combines energy minimization with the level set method.…”

Image Segmentation Adapted for Clinical Settings by Combining Pattern Classification and Level Sets

“…Coupled level set on the other hand uses one level set function to represent each object. But it is not only slow but it also suffered from the problem of placement of initial curve, a common problem that exists in numerical minimization when functions are non-convex the numerical results may depend on the choice of the initial curves [6,15].…”

“…Later a hierarchical scheme was used to extend this method to segment multiple regions [14]. Very recently Michal et al [6] proposed a hierarchical volumetric segmentation which combines Chan and Vese method and geodesic active contour [1] with edge alignment for volumetric segmentation. To apply level set methods to real time applications, LeFohn [8] has translated level set techniques to graphic cards and run in nearly real time.…”

Geodesic Active Regions Using Non-parametric Statistical Regional Description and Their Application to Aneurysm Segmentation from CTA

“…In places with weak gradient, the region-based information drives the evolution of the active contour providing more robust segmentation. Previous attempts for the inclusion of statistical region-based information into the deformable model has shown promising results in the segmentation of brain aneurysms in 3DRA and CTA data [3,8].…”