Hans Frimmel

Computerized detection of colorectal masses in CT colonography based on fuzzy merging and wall-thickening analysis.

In recent years, several computer-aided detection (CAD) schemes have been developed for the detection of polyps in CT colonography (CTC). However, few studies have addressed the problem of computerized detection of colorectal masses in CTC. This is mostly because masses are considered to be well visualized by a radiologist because of their size and invasiveness. Nevertheless, the automated detection of masses would naturally complement the automated detection of polyps in CTC and would produce a more comprehensive computer aid to radiologists. Therefore, in this study, we identified some of the problems involved with the computerized detection of masses, and we developed a scheme for the computerized detection of masses that can be integrated into a CAD scheme for the detection of polyps. The performance of the mass detection scheme was evaluated by application to clinical CTC data sets. CTC was performed on 82 patients with helical CT scanners and reconstruction intervals of 1.0-5.0 mm in the supine and prone positions. Fourteen patients (17%) had a total of 14 masses of 30-50 mm, and sixteen patients (20%) had a total of 30 polyps 5-25 mm in diameter. Four patients had both polyps and masses. Fifty-six of the patients (68%) were normal. The CTC data were interpolated linearly to yield isotropic data sets, and the colon was extracted by use of a knowledge-guided segmentation technique. Two methods, fuzzy merging and wall-thickening analysis, were developed for the detection of masses. The fuzzy merging method detected masses with a significant intraluminal component by separating the initial CAD detections of locally cap-like shapes within the colonic wall into mass candidates and polyp candidates. The wall-thickening analysis detected nonintraluminal masses by searching the colonic wall for abnormal thickening. The final regions of the mass candidates were extracted by use of a level set method based on a fast marching algorithm. False-positive (FP) detections were reduced by a quadratic discriminant classifier. The performance of the scheme was evaluated by use of a leave-one-out (round-robin) method with by-patient elimination. All but one of the 14 masses, which was partially cut off from the CTC data set in both supine and prone positions, were detected. The fuzzy merging method detected 11 of the masses, and the wall-thickening analysis detected 3 of the masses including all nonintraluminal masses. In combination, the two methods detected 13 of the 14 masses with 0.21 FPs per patient on average based on the leave-one-out evaluation. Most FPs were generated by extrinsic compression of the colonic wall that would be recognized easily and quickly by a radiologist. The mass detection methods did not affect the result of the polyp detection. The results indicate that the scheme is potentially useful in providing a high-performance CAD scheme for the detection of colorectal neoplasms in CTC.

Three-dimensional computer reconstruction of prostate cancer from radical prostatectomy specimens: evaluation of the model by core biopsy simulation

OBJECTIVES A technique was developed for three-dimensional (3D) modeling of prostate cancer and transrectal biopsies. To test the model, the cancer yield of a simulated 10-biopsy protocol was compared with a simulated sextant protocol and with preoperative biopsies regarding cancer detection and correlation with tumor volume. METHODS Transrectal ultrasound-guided core biopsies were taken from 81 men according to a standardized 10-biopsy protocol that included sextant biopsies. The patients underwent radical prostatectomy and specimens were step-sectioned and whole-mounted. Cancer and the prostate capsule were outlined on the slides and the regions transferred to a computer software program developed by our group. A 3D volume of each prostate was reconstructed from the sections. Virtual core biopsy needles imitating the positions of the real biopsies were inserted into the prostate and the cancer yield was calculated. Only the standardized positions were considered in this study (ie, additional biopsies from hypoechoic foci were not accounted for). RESULTS Of the cancers detected with 10 standardized virtual biopsies, 24% would have remained undetected with sextant biopsies. The cancer yield of 10 virtual biopsies correlated with the preoperative biopsies (r = 0.64) and with the tumor volume (r = 0.56). A multiple regression analysis showed that the cancer yield of a simulation of 10 biopsies correlated better with tumor volume than did a simulation of sextant biopsies (P = 0.02). CONCLUSIONS We conclude that computer-assisted 3D reconstruction of prostate cancer can be used as a model for evaluation and optimization of biopsy protocols.

Virtual endoscopic visualization of the colon by shape-scale signatures

We developed a new visualization method for virtual endoscopic examination of computed tomographic (CT) colonographic data by use of shape-scale analysis. The method provides each colonic structure of interest with a unique color, thereby facilitating rapid diagnosis of the colon. Two shape features, called the local shape index and curvedness, are used for defining the shape-scale spectrum. When we map the shape index and curvedness values within CT colonographic data to the shape-scale spectrum, specific types of colonic structures are represented by unique characteristic signatures in the spectrum. The characteristic signatures of specific types of lesions can be determined by use of computer-simulated lesions or by use of clinical data sets subjected to a computerized detection scheme. The signatures are used for defining a two-dimensional color map by assignment of a unique color to each signature region. The method was evaluated visually by use of computer-simulated lesions and clinical CT colonographic data sets, as well as by an evaluation of the human observer performance in the detection of polyps without and with the use of the color maps. The results indicate that the coloring of the colon yielded by the shape-scale color maps can be used for differentiating among the chosen colonic structures. Moreover, the results indicate that the use of the shape-scale color maps can improve the performance of radiologists in the detection of polyps in CT colonography.

Fast and robust computation of colon centerline in CT colonography

Although several methods for generating the centerline of a colon from CT colonographic scans have been proposed, in general they are time-consuming and do not take into account that the images of the colon may be of nonoptimal quality, with collapsed regions, and stool within the colon. Furthermore, the colonic lumen or wall, which is often used as a basis for computation of a centerline, is not always precisely segmented. In this study, we have developed an algorithm for computation of a colon centerline that is fast compared to the centerline algorithms presented in the reviewed literature, and that relies little on a complete colon segments identification. The proposed algorithm first extracts local maxima in a distance map of a segmented colonic lumen. The maxima are considered to be nodes in a set of graphs, and are iteratively linked together, based on a set of connection criteria, giving a minimum distance spanning tree. The connection criteria are computed from the distance from object boundary, the Euclidean distance between nodes and the voxel values on the pathway between pairs of nodes. After the last iteration, redundant branches are removed and end segments are recovered for each remaining graph. A subset of the initial maxima is used for distinguishing between the colon and noncolonic centerline segments among the set of graphs, giving the final centerline representation. A phantom study showed that, with respect to phantom variations, the algorithm achieved nearly constant computation time (2.3-2.9 s) except for the most extreme setting (20.2 s). The algorithm successfully found all, or most of, the centerline (93% - 100%). Displacement from optimum varied with colon diameter (1.2-6.6 mm). By use of 40 CT colonographic scans, the computer-generated centerlines were compared with the centerlines generated by three radiologists. The similarity was measured based on percent coverage and average displacement. The computer-generated centerlines, when compared with human-generated centerlines, had approximately the same displacement as when the human-generated centerlines were compared among each other (3.8 mm versus 4.0 mm). The coverage of the computer-generated centerlines was slightly less than that of the human-generated centerlines (92% versus 94%). The 40 centerlines were, on average, computed in 10.5 seconds, including computation time for the distance transform, with an Intel Pentium-based 800 MHz computer, as compared with 12-17 seconds or more (excluding computation time for the distance transform needed) per centerline as reported in other studies.

Three-dimensional modeling of biopsy protocols for localized prostate cancer

Prostate cancer is the most common malignant tumor in American men, yet only a small percentage of men will develop clinically significant disease. Needle core biopsies are used to confirm the presence of cancer prior to surgery. While needle core biopsies have shown some ability to predict tumor volume and grade in prostatectomy specimens, for the individual patient they are neither sensitive nor specific enough to guide therapy. In this paper, we describe a system for simulating needle biopsies on three-dimensional models of cancerous prostates reconstructed from serial sections. First we segment the serial sections, delineating tumors and landmarks. Next, we register the sections using a color-merging scheme, and reconstruct the three-dimensional model using modified-shape-based interpolation. The resulting volume can be rendered, and simulated needle core biopsies can be taken from the reconstructed model. We use our system to simulate two different biopsy protocols on a reconstructed prostate specimen.

Fast level‐set based image segmentation using coherent propagation

PURPOSE The level-set method is known to require long computation time for 3D image segmentation, which limits its usage in clinical workflow. The goal of this study was to develop a fast level-set algorithm based on the coherent propagation method and explore its character using clinical datasets. METHODS The coherent propagation algorithm allows level set functions to converge faster by forcing the contour to move monotonically according to a predicted developing trend. Repeated temporary backwards propagation, caused by noise or numerical errors, is then avoided. It also makes it possible to detect local convergence, so that the parts of the boundary that have reached their final position can be excluded in subsequent iterations, thus reducing computation time. To compensate for the overshoot error, forward and backward coherent propagation is repeated periodically. This can result in fluctuations of great magnitude in parts of the contour. In this paper, a new gradual convergence scheme using a damping factor is proposed to address this problem. The new algorithm is also generalized to non-narrow band cases. Finally, the coherent propagation approach is combined with a new distance-regularized level set, which eliminates the needs of reinitialization of the distance. RESULTS Compared with the sparse field method implemented in the widely available ITKSnap software, the proposed algorithm is about 10 times faster when used for brain segmentation and about 100 times faster for aorta segmentation. Using a multiresolution approach, the new method achieved 50 times speed-up in liver segmentation. The Dice coefficient between the proposed method and the sparse field method is above 99% in most cases. CONCLUSIONS A generalized coherent propagation algorithm for level set evolution yielded substantial improvement in processing time with both synthetic datasets and medical images.

New high-performance CAD scheme for the detection of polyps in CT colonography

We developed a new method for automated detection of colonic polyps in CT colonography. The colon is extracted from CT images by use of a centerline-based colon segmentation method. Polyp candidates are detected by use of hysteresis thresholding and fuzzy merging. The regions of the polyp candidates are segmented by use of conditional morphological dilation. False-positive polyp candidates are reduced by a region-based supine-prone correspondence method and by a Bayesian neural network with shape and texture features. To evaluate the method, CT colonography was performed for 121 patients with standard technique and single- and multi-detector helical scanners by use of 2.5-5.0 mm collimations, 1.0-5.0 mm reconstruction intervals, and 60-100 mA tube currents. Twenty-eight patients had a total of 42 polyps: 22 polyps were 5-10 mm, and 20 polyps were 11-25 mm in size. A leave-one-out evaluation of the CAD scheme with by-patient elimination yielded 93% by-polyp and by-patient detection sensitivities with 2.0 false-positive detections per data set on average. The average computation time was 4 minutes per data set. The results indicate that the CAD scheme may be useful in improving the performance of computer-aided detection for colon cancer in a clinical screening setting.

Biopsy protocol stability in a three-dimensional model of prostate cancer : Changes in cancer yield after adjustment of biopsy positions

OBJECTIVES Transrectal ultrasound-guided prostate biopsies are often taken according to a systematic, standardized schedule. The diagnostic stability of this system was evaluated by moving the biopsies in a three-dimensional (3D) model. METHODS A computerized 3D reconstruction was made from each of 75 radical prostatectomy specimens. Simulated core biopsies imitated a standardized 10-biopsy protocol, including sextant biopsies. In total, 30,000 biopsies were generated by moving the standardized biopsies 1, 2, 3, and 4 mm (parallel needle shifts) or 5 degrees, 10 degrees, 15 degrees, and 20 degrees(rotation of the needle tip) in a random direction. RESULTS The diagnosis of the individual biopsy changed from cancer to benign or vice versa in 4.9% to 1 5.7% after 1 to 4-mm parallel needle shifts and 2.0% to 7.5% after 5 degrees to 20 degrees rotations. The corresponding figures for the final diagnosis of the 10-biopsy set were 0.8% to 9.6% and 0.5% to 3.2%. Transition zone biopsies containing cancer changed to benign more often than the other biopsies (P <0.001). Parallel needle shifts of 2 mm changed the diagnosis more often than the 15 degrees rotation (9.4% and 5.9%, respectively, P <0.001), although conveying the same overall needle shift. CONCLUSIONS The cancer yield of prostate biopsies is influenced even by small changes in needle positions. The transition zone biopsies are most likely to change from cancer to benign when moved. Changing the insertion point of the needle has a higher impact on cancer yield than rotating the tip.

Fast and robust method to compute colon centerline in CT colonography

We developed a method for generating the centerline of a colon in CT Colonography that is computationally fast, and robust to collapsed regions. Patients underwent CT Colonography after standard pre-colonoscopy cleansing. The colonic lumen was segmented using an existing anatomy-based approach, and a distance map of the colonic lumen was computed using a distance transform. The centerline was computed as follows: Local maxima representative for the centerline were sparsely extracted from the distance map. Iteratively, each pair of maxima satisfying a set of connection criteria were connected, creating a graph-like structure containing a main centerline with additional branches. Branches were later removed and the resulting centerline was stored. Centerlines of the colon were computed, and also manually and independently drawn by two radiologists, for 33 CT Colonographic data sets. The data sets were chosen to give a wide spectrum of colons, ranging from cases with good segmentation and extension to cases with collapsed regions and numerous extra-colonic components such as small bowel. On average, 94% of the human-generated centerlines were correctly identified by the computer-generated centerlines. The average displacement between the human- and computer-generated centerlines was 4.0 mm. Average centerline computation time was less than 4 seconds.

Fast vascular skeleton extraction algorithm

First step in fully automatic high-speed vascular tree segmentation.Vascular skeleton is extracted directly from CT volume without preprocessing.Transferring the problem to a lower dimensional space improves the speed.Automatically adapts to CT value variation between patients.Automatically detects abnormal situations (calcification, implants). Vascular diseases are a common cause of death, particularly in developed countries. Computerized image analysis tools play a potentially important role in diagnosing and quantifying vascular pathologies. Given the size and complexity of modern angiographic data acquisition, fast, automatic and accurate vascular segmentation is a challenging task.In this paper we introduce a fully automatic high-speed vascular skeleton extraction algorithm that is intended as a first step in a complete vascular tree segmentation program. The method takes a 3D unprocessed Computed Tomography Angiography (CTA) scan as input and produces a graph in which the nodes are centrally located artery voxels and the edges represent connections between them. The algorithm works in two passes where the first pass is designed to extract the skeleton of large arteries and the second pass focuses on smaller vascular structures. Each pass consists of three main steps. The first step sets proper parameters automatically using Gaussian curve fitting. In the second step different filters are applied to detect voxels nodes that are part of arteries. In the last step the nodes are connected in order to obtain a continuous centerline tree for the entire vasculature. Structures found, that do not belong to the arteries, are removed in a final anatomy-based analysis. The proposed method is computationally efficient with an average execution time of 29s and has been tested on a set of CTA scans of the lower limbs achieving an average overlap rate of 97% and an average detection rate of 71%.