We propose a method for automatically correcting the spherical topology of any segmentation under any digital connectivity. A multiple region growing process, concurrently acting on the foreground and the background, ...
详细信息
ISBN:
(数字)9783540399032
ISBN:
(纸本)3540204644
We propose a method for automatically correcting the spherical topology of any segmentation under any digital connectivity. A multiple region growing process, concurrently acting on the foreground and the background, divides the segmentation into connected components and successive minimum cost decisions guarantee convergence to correct spherical topology. In contrast to existing procedures that suppose specific initial segmentation (full connectivity, no cavities...) and are designed for a particular task (cortical representation), no assumption is made on the initial image. Our method applied to subcortical segmentations allows us to correct the topology of fourteen non-cortical structures in less than a minute.
Finding point correspondence in anatomical images is a key step in shape analysis and deformable registration. This paper proposes an automatic correspondence detection algorithm using wavelet-based attribute vectors ...
详细信息
ISBN:
(数字)9783540399032
ISBN:
(纸本)3540204644
Finding point correspondence in anatomical images is a key step in shape analysis and deformable registration. This paper proposes an automatic correspondence detection algorithm using wavelet-based attribute vectors defined on every image voxel. The attribute vector reflects the anatomical characteristics in a large neighborhood around the respective voxel. It plays the role of a morphological signature for each voxel and is therefore made as distinctive as possible. Correspondence is then determined via similarity of attribute vectors. Experiments with brain MR images show that the algorithm performs at least as well as human experts, even for complex cortical structures.
Non-rigid volumetric registration has many applications, including inter-patient image fusion, motion quantification, and automatic atlas-based segmentation. Computation time is often a limiting factor in using curren...
详细信息
ISBN:
(纸本)3540204644
Non-rigid volumetric registration has many applications, including inter-patient image fusion, motion quantification, and automatic atlas-based segmentation. Computation time is often a limiting factor in using current methods in clinical environments. Minimizing computation time requires both the internal and the external force updates to be as efficient as possible. In this article, we concentrate on the choice of the external force function. We compare different methods based on optical flow and propose a new correlation-based external force function. In addition, we propose an acceleration technique and study its effect on image quality and the speed of convergence. The results indicate that the acceleration technique improves both the speed and quality, and increases the stability of all the external force methods considered.
During endoscopic operations the surgeon works without direct visual contact to the operation area. The image of the operation situs is displayed on a monitor. Currently, only hardware based image enhancement methods ...
详细信息
ISBN:
(数字)9783540399032
ISBN:
(纸本)3540204644
During endoscopic operations the surgeon works without direct visual contact to the operation area. The image of the operation situs is displayed on a monitor. Currently, only hardware based image enhancement methods are used (e. g., white balance) and often only once at the beginning of an operation. In this contribution we describe a system for real-time endoscopic image enhancement: a typical video-endoscopic system was extended by a computer and a second monitor. Thus the enhanced and the original image can be displayed at the same time. The implemented image enhancement methods (temporal filtering, undistortion and color normalization) were evaluated by 14 surgeons and the results showed that the enhanced images were preferred. The system was already used during a real operation.
Accuracy and,precision of inserting joint implants is essential for successful long-term prosthetic function. Clinical and radiographic assessment of Oxford unicompartmental knee arthroplasty using 3computer-assisted...
详细信息
ISBN:
(数字)9783540399032
ISBN:
(纸本)3540204644
Accuracy and,precision of inserting joint implants is essential for successful long-term prosthetic function. Clinical and radiographic assessment of Oxford unicompartmental knee arthroplasty using 3computer-assisted techniques (CT based, "morphing" and fluoro-based) was undertaken on 17 patients. They were followed for at least one year postoperatively. Clinical outcome assessments: Knee Society Scores, WOMAC and SF-36 all indicated good to excellent results. All methods worked well with the manufacture tolerance of 10degrees in both planes. However, CT based and fluoro-based techniques were significantly better than the morphing technique.
We propose a multimodal free-form registration algorithm that matches voxel class labels rather than image intensities. Individual voxels are displaced such as to minimize the Kullback-Leibler distance between the act...
详细信息
ISBN:
(数字)9783540399032
ISBN:
(纸本)3540204644
We propose a multimodal free-form registration algorithm that matches voxel class labels rather than image intensities. Individual voxels are displaced such as to minimize the Kullback-Leibler distance between the actual and ideal joint probability distribution of voxel class labels, which are assigned to each image individually by a previous segmentation process. We evaluate the performance of the method for inter-subject brain registration with simulated deformations, using a viscous fluid model for regularization. The root mean square difference between recovered and ground truth deformations is smaller than 1 voxel.
Hybrid frameworks combining region-based and boundary-based segmentation methods have been used in 3D medicalimage segmentation applications. In this paper we propose a hybrid 3D segmentation framework which combines...
详细信息
ISBN:
(数字)9783540399032
ISBN:
(纸本)3540204644
Hybrid frameworks combining region-based and boundary-based segmentation methods have been used in 3D medicalimage segmentation applications. In this paper we propose a hybrid 3D segmentation framework which combines Gibbs models, marching cubes and deformable models. We use Gibbs models to create 3D binary masks of the object. Then we use the marching cubes method to initialize a deformable model based on the mask. The deformable model will fit to the object surface driven by the gradient information in the original image. The deformation result will then be used to update the parameters of Gibbs models. These methods will work recursively to achieve a final segmentation. By using the marching cubes method, we succeed in improving the accurancy and efficiency of 3D segmentation. We validate our method by comparing the, segmentation result with expert manual segmentation, the results show that high quality segmentation can be achieved with computational efficiency.
In this paper we introduce user-defined segmentation constraints within the level set methods. Snake-driven methods are powerful and widely explored techniques for object extraction. Level set representations is a mat...
详细信息
ISBN:
(数字)9783540399032
ISBN:
(纸本)3540204644
In this paper we introduce user-defined segmentation constraints within the level set methods. Snake-driven methods are powerful and widely explored techniques for object extraction. Level set representations is a mathematical framework technique to implement such methods. This formulation is implicit, intrinsic and parameter/topology free. Introducing shape-driven knowledge within the level set method for segmentation is a recently explored topic. User interactive constraints are more simplistic forms of prior shape knowledge. To this end, we propose a simple formulation that converts user interaction to objective function terms that aim to guide the segmentation solution through the user edits.
We combine a new user initialization process with a B-spline snake to create a model with the properties of a deformable template. This 'template' snake can be constrained by its control polygon and is initial...
详细信息
ISBN:
(数字)9783540399032
ISBN:
(纸本)3540204644
We combine a new user initialization process with a B-spline snake to create a model with the properties of a deformable template. This 'template' snake can be constrained by its control polygon and is initially extremely close to, and similar in shape to, the target anatomical structure. The initialization process acts as almost a pre-segmentation and labelling step, making the snake's task much simpler and hence more likely to succeed in noisy images without subsequent user editing. By imposing an order on the initialization process, the user is able to transfer knowledge of global shape, symmetry, landmark position etc. to the model. We apply our snake to the segmentation of 2D medicalimages.
Our method is composed of three steps: The 3D image is projected along an initial direction, and the needle is segmented in the projected image. Using the projection direction and detected 2D needle direction, a plane...
详细信息
ISBN:
(数字)9783540399032
ISBN:
(纸本)3540204644
Our method is composed of three steps: The 3D image is projected along an initial direction, and the needle is segmented in the projected image. Using the projection direction and detected 2D needle direction, a plane containing the needle, called the needle plane, is determined. Secondly, the 3D image is projected in the direction perpendicular to the normal of the needle plane and Step 1 is repeated. If the needle direction in the projected image is horizontal, the needle plane is correct, otherwise Steps 1 & 2 are repeated until a correct needle plane has been found. Thirdly, the 3D image is projected along the normal direction of the needle plane. In this projected image, we determine the endpoints of the needle in the projection coordinates. Then, using the relationship between the projection coordinates and the 3D image coordinates, we determine the end points of the needle in the 3D US image. Experiments with agar and turkey phantom 3D US images demonstrated that our method can segment the needle from 3D US images accurately and robustly with a speed of 13 fps on a 1.8GHz PC computer.
暂无评论