Yuanzhi Cheng

Automatic segmentation technique for acetabulum and femoral head in CT images

Abstract Segmentation of the femoral head and proximal acetabulum from three dimensional (3D) CT data is essential for patient specific planning and simulation of hip surgery whereas it still remains challenging due to deformed shapes and extremely narrow inter-bone regions. In this paper, we present an accurate, automatic and fast approach for simultaneous segmentation of the femoral head and proximal acetabulum in the hip joint from 3D CT data. First valley-emphasized image is constructed from original images so that valleys stand out in high relief and initial thresholding segmentation is performed to divide the image set into bone (femoral head and acetabulum) and non-bone classes. It is employed as an initial boundary of the femoral head and acetabulum for further processing in the segmentation procedures. In the subsequent iterative process, the bone regions are further segmented with consideration of the narrow joint space, the neighborhood information and the partial volume effect. Finally, the segmented bone boundaries are corrected based on the normal direction of vertices of the 3D bone surface. Evaluation of the method is performed on the 110 hips including pathologies. Experimental results indicate that our method rapidly leads to very accurate segmentations of the femoral head and acetabulum in the hip joint and can be applied as a tool in the clinical practice.

Accurate Vessel Segmentation With Constrained B-Snake

We describe an active contour framework with accurate shape and size constraints on the vessel cross-sectional planes to produce the vessel segmentation. It starts with a multiscale vessel axis tracing in a 3D computed tomography (CT) data, followed by vessel boundary delineation on the cross-sectional planes derived from the extracted axis. The vessel boundary surface is deformed under constrained movements on the cross sections and is voxelized to produce the final vascular segmentation. The novelty of this paper lies in the accurate contour point detection of thin vessels based on the CT scanning model, in the efficient implementation of missing contour points in the problematic regions and in the active contour model with accurate shape and size constraints. The main advantage of our framework is that it avoids disconnected and incomplete segmentation of the vessels in the problematic regions that contain touching vessels (vessels in close proximity to each other), diseased portions (pathologic structure attached to a vessel), and thin vessels. It is particularly suitable for accurate segmentation of thin and low contrast vessels. Our method is evaluated and demonstrated on CT data sets from our partner site, and its results are compared with three related methods. Our method is also tested on two publicly available databases and its results are compared with the recently published method. The applicability of the proposed method to some challenging clinical problems, the segmentation of the vessels in the problematic regions, is demonstrated with good results on both quantitative and qualitative experimentations; our segmentation algorithm can delineate vessel boundaries that have level of variability similar to those obtained manually.

Low-Rank and Sparse Decomposition Based Shape Model and Probabilistic Atlas for Automatic Pathological Organ Segmentation

HIGHLIGHTSWe propose an automatic pathological organ CT segmentation based on ASM framework.A low‐rank and sparse decomposition (LRSD) based shape prior model is presented.A shape initialization method using LRSD‐based probabilistic atlas is introduced.A hierarchical ASM search strategy is developed to make the framework efficient.The method is successfully applied to segment pathological liver and right lung. ABSTRACT One major limiting factor that prevents the accurate delineation of human organs has been the presence of severe pathology and pathology affecting organ borders. Overcoming these limitations is exactly what we are concerned in this study. We propose an automatic method for accurate and robust pathological organ segmentation from CT images. The method is grounded in the active shape model (ASM) framework. It leverages techniques from low‐rank and sparse decomposition (LRSD) theory to robustly recover a subspace from grossly corrupted data. We first present a population‐specific LRSD‐based shape prior model, called LRSD‐SM, to handle non‐Gaussian gross errors caused by weak and misleading appearance cues of large lesions, complex shape variations, and poor adaptation to the finer local details in a unified framework. For the shape model initialization, we introduce a method based on patient‐specific LRSD‐based probabilistic atlas (PA), called LRSD‐PA, to deal with large errors in atlas‐to‐target registration and low likelihood of the target organ. Furthermore, to make our segmentation framework more efficient and robust against local minima, we develop a hierarchical ASM search strategy. Our method is tested on the SLIVER07 database for liver segmentation competition, and ranks 3rd in all the published state‐of‐the‐art automatic methods. Our method is also evaluated on some pathological organs (pathological liver and right lung) from 95 clinical CT scans and its results are compared with the three closely related methods. The applicability of the proposed method to segmentation of the various pathological organs (including some highly severe cases) is demonstrated with good results on both quantitative and qualitative experimentation; our segmentation algorithm can delineate organ boundaries that reach a level of accuracy comparable with those of human raters.

Accuracy Limits for the Thickness Measurement of the Hip Joint Cartilage in 3-D MR Images: Simulation and Validation

This paper describes a theoretical simulation method for ascertaining the inherent limits on the accuracy of thickness measurement of hip joint cartilage in 3-D MR images. This method can specify where and how thickness can be measured with sufficient accuracy under the certain MR imaging conditions. In the numerical simulation, we present a mathematical model for two adjacent sheet structures separated by a small distance, which simulated the femoral and acetabular cartilage and the joint space width in the hip joint; moreover, we perform the numerical simulation of MR imaging and postprocessing for thickness measurement. We especially focused on the effects of voxel anisotropy in MR imaging with variable orientation of cartilage surface and different joint space width. Also, thickness measurement is performed in MR imaging with isotropic voxel. The results from MR data with isotropic voxels show that accurate measurement of cartilage thickness at location of measured values of the hip joint space width and the cartilage thickness being two times as large as the voxel size or above should be possible. The simulation method is validated by comparison with the actual results obtained from the experiments using three phantoms, five normal cadaver hip specimens, and nine patients with osteoarthritis.

A technique for visualization and mapping of local cartilage thickness changes in MR images of osteoarthritic knee.

PURPOSE The aim of this paper is to describe a technique for the visualization and mapping of focal, local cartilage thickness changes over time in magnetic resonance images of osteoarthritic knee. METHODS Magnetic resonance imaging was performed in 25 fresh frozen pig knee joints and 15 knees of patients with borderline to mild osteoarthritis (51.2±6.3 years). Cartilage and corresponding bone structures were extracted by semi-automatic segmentation. Each point in the bone surface which was part of the bone-cartilage interface was assigned a cartilage thickness value. Cartilage thicknesses were computed for each point in the bone-cartilage interfaces and transferred to the bone surfaces. Moreover, we developed a three dimensional registration method for the identification of anatomically corresponding points of the bone surface to quantify local cartilage thickness changes. One of the main advantages of our method compared to other studies in the field of registration is a global optimization algorithm that does not require any initialization. RESULTS AND CONCLUSION The registration accuracy was 0.93±0.05 mm (less than a voxel of magnetic resonance data). Local cartilage thickness changes were seen as having follow-up clinical study for detecting local changes in cartilage thickness. Experiment results suggest that our method was sufficiently accurate and effective for monitoring knee joint diseases.

Surface-based rigid registration using a global optimization algorithm for assessment of MRI knee cartilage thickness changes

Abstract Registration methods have become an important tool in many medical applications. Existing methods require a good initial estimation (transformation) in order to find a global solution, i.e., if the initial estimation is far from the actual solution, incorrect solution or mismatching is very likely. In contrast, this paper presents a novel approach for globally solving the three dimensional (3D) rigid registration problem. The registration is grounded on a mathematical theory—Lipschitz optimization. It achieves a guaranteed global optimality with a rough initial estimation (e.g., even a random guess). Moreover, Munkres assignment algorithm is used to find the point correspondences. It applies the distance matrix to find an optimal correspondence. Our method is evaluated and demonstrated on MR images from porcine knees and human knees. Compared with state-of-the-art methods, the proposed technique is more robust, more accurate to perform point to point comparisons of knee cartilage thickness values for follow-up studies on the same subject.

Greedy algorithm based deformable simplex meshes using gradient vector flow as external energy

Deformable models have been quite popular in medical image analysis, particularly in image segmentation. However, when applied to 3D volumetric data, their high computational cost can be a problem. In this paper, we describe a new efficient 3D segmentation method based on deformable simplex meshes. The greedy algorithm, which has proven more computational efficient and robust than physics-based method, is employed to perform the shape deformation. Generalized gradient vector flow (GGVF) field is a classical external force for physics-based deformable models. We adapt it for greedy algorithm as external energy to overcome the main issues of the traditional external energy (i.e., sensitivity to shape initialization and poor convergence to the long and thin boundary concavities). Results of applying our method to both synthetic and clinical images are presented to illustrate the accuracy and robustness of our proposed method.

Automatic centerline detection of small three-dimensional vessel structures

Abstract. Vessel centerline detection is very important in many medical applications. In the noise and low-contrast regions, most existing methods may only produce an incomplete and disconnected extraction of the vessel centerline if no user guidance is provided. A robust and automatic method is described for extraction of the vessel centerline. First, we perform small vessel enhancement by processing with a set of line detection filters, corresponding to the 13 orientations; for each voxel, the highest filter response is kept and added to the image. Second, we extract vessel centerline segment candidates by a thinning algorithm. Finally, a global optimization algorithm is employed for grouping and selecting vessel centerline segments. We validate the proposed method quantitatively on a number of synthetic data sets, the liver artery and lung vessel. Comparisons are made with two state-of-the-art vessel centerline extraction methods and manual extraction. The experiments show that our method is more accurate and robust that these state-of-the-art methods and is, therefore, more suited for automatic vessel centerline extraction.

Segmentation of the hip joint in CT volumes using adaptive thresholding classification and normal direction correction

Segmentation of the pelvis and proximal femur in computed tomography (CT) volumes is a prerequisite of patient specific planning and simulation for hip surgery. Existing methods do not perform well due to bone disease and technical limitations of CT imaging. In this paper, an accurate framework for segmenting bone in the hip joint is presented. Our approach begins with valley-emphasized image construction using morphological operations so that valleys stand out in high relief, and then, an initial segmentation with optimal threshold is performed to divide the dataset into bone and non-bone regions. Subsequently, bone regions are reclassified based on 3D iterative adaptive thresholding with consideration of the partial volume effect and the spatial information. Finally, we refine the rough bone boundaries based on the normal direction of vertices of the 3D bone surface. Our segmentation approach is automatic and robust. Its performance is evaluated on 35 datasets consisting of 70 hip joints with a status ranging from healthy to severe osteoarthritis and the results have proved to be very successful.

Sparse Representation-Based Deformation Model for Atlas-Based Segmentation of Liver CT Images

Liver segmentation in computed tomography (CT) images is a fundamental step for various computer-assisted clinical applications. However, automatic liver segmentation from CT images is still a challenging task. In this paper, we propose a novel non-parametric sparse representation-based deformation model (SRDM) for atlas-based liver segmentation framework using nonrigid registration based on free-form deformations (FFDs) model. Specifically, during atlas-based segmentation procedure, our proposed SRDM provides a regularization for the resulting deformation that maps the atlas to the space of the target image, constraining it to be a sparse linear combination of existing training deformations in a deformation repository. We evaluated our proposed method based on a set of 30 contrast-enhanced abdominal CT images, resulting in superior performance when compared to state-of-the-art atlas-based segmentation methods.