Karl Krissian

Standardized evaluation methodology and reference database for evaluating coronary artery centerline extraction algorithms.

Efficiently obtaining a reliable coronary artery centerline from computed tomography angiography data is relevant in clinical practice. Whereas numerous methods have been presented for this purpose, up to now no standardized evaluation methodology has been published to reliably evaluate and compare the performance of the existing or newly developed coronary artery centerline extraction algorithms. This paper describes a standardized evaluation methodology and reference database for the quantitative evaluation of coronary artery centerline extraction algorithms. The contribution of this work is fourfold: (1) a method is described to create a consensus centerline with multiple observers, (2) well-defined measures are presented for the evaluation of coronary artery centerline extraction algorithms, (3) a database containing 32 cardiac CTA datasets with corresponding reference standard is described and made available, and (4) 13 coronary artery centerline extraction algorithms, implemented by different research groups, are quantitatively evaluated and compared. The presented evaluation framework is made available to the medical imaging community for benchmarking existing or newly developed coronary centerline extraction algorithms.

Evaluation framework for carotid bifurcation lumen segmentation and stenosis grading.

This paper describes an evaluation framework that allows a standardized and objective quantitative comparison of carotid artery lumen segmentation and stenosis grading algorithms. We describe the data repository comprising 56 multi-center, multi-vendor CTA datasets, their acquisition, the creation of the reference standard and the evaluation measures. This framework has been introduced at the MICCAI 2009 workshop 3D Segmentation in the Clinic: A Grand Challenge III, and we compare the results of eight teams that participated. These results show that automated segmentation of the vessel lumen is possible with a precision that is comparable to manual annotation. The framework is open for new submissions through the website http://cls2009.bigr.nl.

Accurate subpixel edge location based on partial area effect

The estimation of edge features, such as subpixel position, orientation, curvature and change in intensity at both sides of the edge, from the computation of the gradient vector in each pixel is usually inexact, even in ideal images. In this paper, we present a new edge detector based on an edge and acquisition model derived from the partial area effect, which does not assume continuity in the image values. The main goal of this method consists in achieving a highly accurate extraction of the position, orientation, curvature and contrast of the edges, even in difficult conditions, such as noisy images, blurred edges, low contrast areas or very close contours. For this purpose, we first analyze the influence of perfectly straight or circular edges in the surrounding region, in such a way that, when these conditions are fulfilled, the features can exactly be determined. Afterward, we extend it to more realistic situations considering how adverse conditions can be tackled and presenting an iterative scheme for improving the results. We have tested this method in real as well as in sets of synthetic images with extremely difficult edges, and in both cases a highly accurate characterization has been achieved.

Segmentation and reconstruction of vascular structures for 3D real-time simulation

We propose a technique to obtain accurate and smooth surfaces of patient specific vascular structures, using two steps: segmentation and reconstruction. The first step provides accurate and smooth centerlines of the vessels, together with cross section orientations and cross section fitting. The initial centerlines are obtained from a homotopic thinning of the vessels segmented using a level set method. In addition to circle fitting, an iterative scheme fitting ellipses to the cross sections and correcting the centerline positions is proposed, leading to a strong improvement of the cross section orientations and of the location of the centerlines. The second step consists of reconstructing the surface based on this data, by generating a set of topologically preserved quadrilateral patches of branching tubular structures. It improves Felkels meshing method (Felkel et al., 2004) by: allowing a vessel to have multiple parents and children, reducing undersampling artifacts, and adapting the cross section distribution. Experiments, on phantom and real datasets, show that the proposed technique reaches a good balance in terms of smoothness, number of triangles, and distance error. This technique can be applied in interventional radiology simulations, virtual endoscopy and in reconstruction of smooth and accurate three-dimensional models for use in simulation.

Symmetric optical flow

One of the main technique used to recover motion analysis from two images or to register them is variational optical flow, where the pixels of one image are matched to the pixels of the second image by minimizing an energy functional. In the standard formulation of variational optical flow, the estimated motion vector field depends on the reference image and is asymmetric. However, in most application the solution should be independent of the reference image. Only few symmetrical formulations of the optical flow has been proposed in the literature, where the solution is constraint to be symmetric using a combination of the flow in both directions. We propose a new symmetric variational formulation of the optical flow problem, where the flow is naturally symmetric. Results on the Yosemite sequence show an improved accuracy of our symmetric flow with respect to standard optical flow algorithm.

Semi-automatic segmentation and detection of aorta dissection wall in MDCT angiography.

Aorta dissection is a serious vascular disease produced by a rupture of the tunica intima of the vessel wall that can be lethal to the patient. The related diagnosis is strongly based on images, where the multi-detector CT is the most generally used modality. We aim at developing a semi-automatic segmentation tool for aorta dissections, which will isolate the dissection (or flap) from the rest of the vascular structure. The proposed method is based on different stages, the first one being the semi-automatic extraction of the aorta centerline and its main branches, allowing an subsequent automatic segmentation of the outer wall of the aorta, based on a geodesic level set framework. This segmentation is then followed by an extraction the center of the dissected wall as a 3D mesh using an original algorithm based on the zero crossing of two vector fields. Our method has been applied to five datasets from three patients with chronic aortic dissection. The comparison with manually segmented dissections shows an average absolute distance value of about half a voxel. We believe that the proposed method, which tries to solve a problem that has attracted little attention to the medical image processing community, provides a new and interesting tool to isolate the intimal flap that can provide very useful information to the clinician.

Comparison of Two Restoration Techniques in the Context of 3D Medical Imaging

In this paper, we compare two restoration techniques applied to 3D angiographies and to femoral CT scans. The first technique uses a Partial Derivative Equation and the second one is based on an extension of adaptive Wiener filters. We first present each method. Then, we discuss and compare the estimation of the local orientations in 3D images obtained either by the smoothed gradient and the principal curvature directions or by the eigenvectors of the structure tensor. A good estimation of the orientations is essential because it directs the restoration process. Finally, we compare the restored images on both synthetic and real images for the two studied applications.

Evaluation of a semi-automatic segmentation algorithm in 3D intraoperative ultrasound brain angiography.

Abstract In this work, we adapted a semi-automatic segmentation algorithm for vascular structures to extract cerebral blood vessels in the 3D intraoperative contrast-enhanced ultrasound angiographic (3D-iUSA) data of the brain. We quantitatively evaluated the segmentation method with a physical vascular phantom. The geometrical features of the segmentation model generated by the algorithm were compared with the theoretical tube values and manual delineations provided by observers. For a silicon tube with a radius of 2 mm, the results showed that the algorithm overestimated the lumen radii values by about 1 mm, representing one voxel in the 3D-iUSA data. However, the observers were more hindered by noise and artifacts in the data, resulting in a larger overestimation of the tube lumen (twice the reference size). The first results on 3D-iUSA patient data showed that the algorithm could correctly restitute the main vascular segments with realistic geometrical features data, despite noise, artifacts and unclear blood vessel borders. A future aim of this work is to provide neurosurgeons with a visualization tool to navigate through the brain during aneurysm clipping operations.

A new energy-based method for 3D motion estimation of incompressible PIV flows

Motion estimation has many applications in fluid analysis, and a lot of work has been carried out using Particle Image Velocimetry (PIV) to capture and measure the flow motion from sequences of 2D images. Recent technological advances allow capturing 3D PIV sequences of moving particles. In the context of 3D flow motion, the assumption of incompressibility is an important physical property that is satisfied by a large class of problems and experiments. Standard motion estimation techniques in computer vision do not take into account the physical constraints of the flow, which is a very interesting and challenging problem. In this paper, we propose a new variational motion estimation technique which includes the incompressibility of the flow as a constraint to the minimization problem. We analyze, from a theoretical point of view, the influence of this constraint and we design a new numerical algorithm for motion estimation which enforces it. The performance of the proposed technique is evaluated from numerical experiments on synthetic and real data.

A subpixel edge detector applied to aortic dissection detection

The aortic dissection is a disease that can cause a deadly situation, even with a correct treatment. It consists in a rupture of a layer of the aortic artery wall, causing a blood flow inside this rupture, called dissection. The aim of this paper is to contribute to its diagnosis, detecting the dissection edges inside the aorta. A subpixel accuracy edge detector based on the hypothesis of partial volume effect is used, where the intensity of an edge pixel is the sum of the contribution of each color weighted by its relative area inside the pixel. The method uses a floating window centred on the edge pixel and computes the edge features. The accuracy of our method is evaluated on synthetic images of different thickness and noise levels, obtaining an edge detection with a maximal mean error lower than 16 percent of a pixel.