Although analysis of scoliotic deformity is still studied extensively by means of conventional roentgenograms, computer-assisted digital analysis may allow a faster, more accurate and more complete evaluation of the scoliotic spine. In this study, a new computer-assisted measurement method was evaluated. This method uses digital reconstruction images for quantitative analysis of the scoliotic spine. The aim of the current study was to determine the reliability of the computer-assisted measuring method, which was done by establishing coefficients of repeatability for a variety of measurements. Measurements were carried out by five observers on 30 frontal and 10 lateral scoliotic digital reconstruction images. Each image was measured on three separate occasions by placing anatomical vertebral landmarks and drawing lines with a computer pointing device. The computer then calculated a number of geometrical shape parameters from scale calibration, landmarks and lines. The intra- and interobserver results were subjected to an analysis of variance to assess the level of agreement, and the means and standard deviations were calculated. The coefficient of repeatability (CR) was taken to be equal to two standard deviations. The mean intraobserver CR was found to be 3.1 degrees for the Cobb angle on the frontal digital image and 3.3 degrees for the kyphosis Cobb angle on the lateral overview. The mean difference in the intraobserver CR of the Cobb angle between measurements made by placing landmarks and those made by drawing lines was not statistically significant (P>0.05). The mean intraobserver CR for the other parameters can be summarized as follows: for lateral deviation it was 0.8 mm, for axial rotation 4.0 degrees and for length of the spine 3.3 mm. The interobserver bias was negligible. It can be concluded that the reliability of our new method for quantifying geometrical variables on digital reconstruction images is better than measurements on conventional roentgenograms in previously published reports. The presented method is therefore considered to be more accurate for research of spinal deformities and more adequate for clinical management of scoliosis.

A method is introduced to automatically find the coronary axis based on two or more user-defined points, even in the presence of a severe stenosis. The coronary axis is determined by finding a minimum cost path (MCP) in a feature image in which the tubular-like structures are enhanced. The results of the proposed method were compared with manually drawn central axes to estimate the accuracy. In 32 3D TFE-EPI acquisitions of patients and volunteers, 14 right coronary arteries (RCAs), 15 left anterior descending arteries (LADs), and eight left circumflex arteries (LCXs) were manually tracked twice by two operators to determine a reference axis and to assess the interand intra-user variability. On average, the maximum distance to the reference axis, based on only two user-defined points, is less than 1.5 mm; the average distance is around 0.65 mm, which is less than the average in-plane resolution. In the last few years several magnetic resonance angiography (MRA) acquisition methods have been introduced which provide 3D images of the coronary arteries with good signal-to-noise (SNR) and contrast-to-noise (CNR) ratios (1-6). However, because of the tortuous nature of the coronaries, it is not possible to capture a long stretch of vessel in a single plane. Postprocessing is therefore required to obtain a proper visualization of the coronaries in 3D space. Although a large variety of approaches have been proposed to facilitate the diagnosis of vessel segments (e.g., 7-15), relatively few have addressed the problems that occur in the case of a (severe) stenosis or in the presence of image artifacts, in which case there is hardly any image evidence to guide the algorithm. Especially iterative tracking and region-growing procedures experience difficulties in these cases (see Fig. 1a).Determining a minimum cost path (MCP) (e.g., 16 -23) between two or more user-defined points is an alternative way to handle these situations (see Fig. 1b).In this article it is investigated whether the latter approach, in combination with a filter that enhances tubularlike structures, is suitable for tracking the coronary arteries automatically between two or more user-defined points. The result of the MCP approach is an estimate of the central coronary axis, which can serve as input for a subsequent visualization and quantification procedure. The method is applied to a number of 3D MRA images, and the results are compared with those of two human operators to determine the accuracy and applicability of the method. METHODS Image AcquisitionThe images are acquired on a 1.5 Tesla Philips Gyroscan ACS-NT MR scanner, using a navigator gated and corrected ECG triggered ultra-fast 3D interleaved gradient turbo field echo-echo planar imaging (TFE-EPI) sequence, preceded by fat-and muscle-suppressing pulses. For a detailed description of the imaging sequence, see Botnar et al. (6). The images consist of 20 slices of 512 ϫ 512 pixels, and a reconstructed slice thickness of 1.5 mm. The field of view (FOV) is either 360 or 370 mm. A total of 32 image...

A method is presented that aims at finding the central vessel axis in two and three dimensional angiographic images based on a single user defined point. After the vessels in the image are enhanced using a special purpose filter, the operator is asked to point out the vessel of interest. Subsequently, a wave front propagation is started based on the response of the filter. By analyzing the evolution of the wave front, points are retrieved that are very likely to be part of the vessel of interest. These points can either be combined to form a connected structure or to retrieve the minimum cost path to the user defined point. In this paper examples of this approach are given that illustrate the performance of this method in different types of images and in situations where there is no or hardly any image evidence of the vessel at hand.

Abstract. Nowadays, conventional X-ray radiographs are still the images of choice for evaluating spinal deformaties such as scoliosis. However, digital translation reconstruction gives easy access to high quality, digital overview images of the entire spine. This work aims at improving the description of the scoliotic deformity by developing semi-automated tools to assist the extraction of anatomical landmarks (on vertebral bodies and pedicles) and the calculation of deformity quantifying parameters. These tools are currently validated in a clinical setting.

No abstract