Computer-aided, case-based diagnosis of mammographic regions of interest containing microcalcifications.

RATIONALE AND OBJECTIVES: The purpose of this study was to evaluate the effectiveness of a mapped-database diagnostic system in reducing the incidence of benign biopsies and misdiagnosed cancers among mammographic regions of interest (ROIs). MATERIALS AND METHODS: A novel neural network was devised (a) to respond to a query ROI by recommending to biopsy or not to biopsy and (b) to map each ROI in the database as a dot on a computer screen. The network was designed so that clusters in the array of dots help the radiologist to find proved ROIs visually similar to the query ROI. This mapped-database diagnostic system was restricted to ROIs with visible microcalcifications. The neural network was trained with a stored database of 80 biopsy-proved ROIs. RESULTS: Four radiologists acting independently on 100 ROIs recommended biopsies for 18, 15, 28, and 18 benign ROIs and misdiagnosed cancers in 11, 12, 7, and eight ROIs, respectively. Interaction with the mapped-database system reduced the numbers of benign biopsies to 11, eight, 18, and 10 cases and of misdiagnosed cancers to eight, seven, four, and three cases, respectively. Statistical analysis indicated that three radiologists achieved significant improvements at P < or = .02 and the fourth achieved a substantial improvement at P < or = .07. CONCLUSION: By using a mapped database of proved mammographic ROIs containing microcalcifications, radiologists may statistically significantly reduce the numbers of benign biopsies and misdiagnosed cancers.

Biplane X-ray angiograms, intravascular ultrasound, and 3D visualization of coronary vessels.

The technology for determination of the 3D vascular tree and quantitative characterization of the vessel lumen and vessel wall has become available. With this technology, cardiologists will no longer rely primarily on visual inspection of coronary angiograms but use sophisticated modeling techniques combining images from various modalities for the evaluation of coronary artery disease and the effects of treatment. Techniques have been developed which allow the calculation of the imaging geometry and the 3D position of the vessel centerlines of the vascular tree from biplane views without a calibration object, i.e., from the images themselves, removing the awkwardness of moving the patient to obtain 3D information. With the geometry and positional information, techniques for reconstructing the vessel lumen can now be applied that provide more accurate estimates of the area and shape of the vessel lumen. In conjunction with these developments, techniques have been developed for combining information from intravascular ultrasound images with the information obtained from angiography. The combination of these technologies will yield a more comprehensive characterization and understanding of coronary artery disease and should lead to improved and perhaps less invasive patient care.

A visual neural classifier.

A new neural classifier allows visualization of the training set and decision regions, providing benefits for both the designer and the user. We demonstrate the visualization capabilities of this visual neural classifier using synthetic data, and compare the visualization performance to Kohunen's self-organizing map. We show in applications to image segmentation and medical diagnosis that visualization enables a designer to refine the classifier to achieve low error rates and enhances a user's ability to make classifier-assisted decisions.

Reconstructing the 3-D medial axes of coronary arteries in single-view cineangiograms.

Describes a technique for reconstructing the skeletal structure of coronary arteries from a succession of frames of a single-view cineangiogram. The authors use local features in each frame to determine correspondences of arterial segments in successive frames. They define a similarity measure in 2D image space as the change in angular coordinates of corresponding pairs. They use a form of gradient descent to find those depth coordinates that minimize the average deviation of the 3D angular coordinates of all points on the skeleton from the coordinates produced by a 3D scaling transformation. In experiments with software models the reconstruction error was approximately two pixels when the initial guessed reconstruction was as large as 30 pixels.

Flexible mask subtraction for digital angiography.

A flexible-mask algorithm that effectively provides correction for motion artifacts and gray-level variations in digital subtraction angiography (DSA) images is presented. The algorithm makes use of a flexible local registration of the mask with the live image to minimize the main distortions caused by the complex motion of the heart, namely spatial translation, rotation, and nonisotropic scaling. It also reduces the gray-level variations caused by the diffusion of the contrast medium into parts of the heart other than the arteries. It is concluded from experiments on X-ray coronary angiograms that when the background near the arteries is rugged or provides significant interference, flexible mask subtraction offers important improvements in quantitative estimates of the sizes of the arteries. In all the X-ray images used in the experiments, the precision of the method was at the 0.1-pixel level. The computation is extensive. However, some improvements can be achieved by doing the subtraction only at the coronary arteries and the immediate neighborhood instead of the whole image. The algorithm lends itself to implementation by parallel-structured hardware.

Reconstructing the cross sections of coronary arteries from biplane angiograms.

An algorithm that reconstructs the cross sections of the lumens of coronary arteries from two mutually orthogonal X-ray projections is described. The algorithm accommodates the possibility of elliptical, crescent, or star shapes. It represents each biplane projection of a transverse slice of the arterial lumen as a binary-valued image. The single-coordinate moments of these two projection images are equal to those of the slice. Since the cross-coordinate moments of the slice are not available from the projections, an algorithm to estimate these moments based on assumptions of smoothness and connectivity is developed. Once all the missing moments are estimated, the image of the slice can be estimated by inverting these moments, using the uniqueness theorem governing the relation between an image and its moments. Preliminary tests of the algorithm on synthetic data, on hardware phantoms and on a segment of a barium-enhanced in vitro coronary artery are reported.

Estimating the 3D skeletons and transverse areas of coronary arteries from biplane angiograms.

A method for estimating the three-dimensional (3D) skeletons and transverse areas of the lumens of coronary arteries from digital X-ray angiograms is described. The method is based on the use of a 3D generalized cylinder (GC) consisting of a series of 3D elliptical disks transverse to and centered on a 3D skeleton (medial axis) of the coronary arteries. The estimates of the transverse areas are based on a nonlinear least-squares-error estimation technique described by D.W. Marquardt (1963). This method exploits densitometric profiles, boundary estimates, and the orientation of the arterial skeleton in 3-space and includes an automatic artery tracking procedure. It applies an adaptive window to the densitometric profile data that are used in the parameter estimation. Preliminary experimental tests of the procedure on angiograms of in vivo human coronaries and on synthetic images yield encouraging results.

Recursive algorithms for implementing digital image filters.

The B-spline functions are used to develop recursive algorithms for the efficient implementation of two-dimensional linear digital image filters. These filters may be spatially varying. The B-splines are used in a representation of the desired point spread function. We show that this leads to recursive algorithms and hardware implementations which are more efficient than either direct spatial domain filter realizations or FFT implementations. The Z-transform is used to develop a discrete version of Duhamel's theorem. A computer architecture for B-spline image filters is proposed and a complexity analysis and comparison to other approaches is provided.

Computing the skeleton of coronary arteries in cineangiograms.

In this paper we describe techniques to determine the skeleton of coronary arteries and to measure their medically significant geometric properties (such as boundaries and orientations) in cineangiograms. We use these geometric properties for segmentation, i.e., for separating the image of the arteries from the background. The purpose of this segmentation is to produce a structural description of the arterial tree for high-level processing. Our technique is completely automatic, efficient, and reliable. Our technique starts with the detection of skeleton points from a local ridge detector using density profiles of the arteries. Local orientations of skeleton points are obtained directly from the ridge detection or from local edge directions. Boundaries of the arteries are computed from the density profile along a straight line perpendicular to the skeleton. From this, a sequential and raster-scan tracking procedure is applied to obtain segments of the arteries from two scanning directions. The final segments are generated by merging all the segments.

Elimination of nonpivotal plane images from x-ray motion tomograms.

A method for eliminating the contribution of nonpivotal plane images into a conventional X-ray motion tomogram is described. A set of tomograms, each focused on one of distinct parallel planes, are combined to form a blur-free tomogram. From the mathematical analysis of conventional tomography, it is found that the multisection tomography system reduces to a three-dimensional incomplete linear system with angularly limited frequency response. An iterative constrained reconstruction algorithm incorporating the object boundary constraint is used for reconstruction of a blur-free tomogram without a contribution from nonpivotal planes. Some computer simulations were made for verification of the algorithm. Although the algorithm accomplishes the three-dimensional reconstruction of volume rather than image restoration of a single slice plane, the optimal high-pass spatial filter applying to a single tomogram is derived on the way to solving the inverse problem of the multisection tomography system.

Undersampling Errors in Region-of-Interest Tomography.

A detailed error analysis for the dual-sampling region-ofinterest X-ray tomography is presented. Simulation studies are used along with a range of sampling rates to quantitate the amount of sampling errors within the region of interest. It is shown that as the rate of sampling outside the region of interest becomes sparse the amount of sampling errors within the region of interest increases considerably.

A gestalt-guided heuristic boundary follower for x-ray images of lung nodules.

A ``piecewise heuristic'' follower of the boundaries of difficult-to-notice nonoverlapping blobs in digitized images is described. The technique was tested on 324 X-ray images of candidate lung nodules from eight radiographs with good results. This technique is several times faster than earlier detectors of the boundaries of lung nodules, while maintaining good fidelity of the detected boundaries.

Locally trained piecewise linear classifiers.

We describe a versatile technique for designing computer algorithms for separating multiple-dimensional data (feature vectors) into two classes. We refer to these algorithms as classifiers. Our classifiers achieve nearly Bayes-minimum error rates while requiring relatively small amounts of memory. Our design procedure finds a set of close-opposed pairs of clusters of data. From these pairs the procedure generates a piecewise-linear approximation of the Bayes-optimum decision surface. A window training procedure on each linear segment of the approximation provides great flexibility of design over a wide range of class densities. The data consumed in the training of each segment are restricted to just those data lying near that segment, which makes possible the construction of efficient data bases for the training process. Interactive simplification of the classifier is facilitated by an adjacency matrix and an incidence matrix. The adjacency matrix describes the interrelationships of the linear segments {£i}. The incidence matrix describes the interrelationships among the polyhedrons formed by the hyperplanes containing {£i}. We exploit switching theory to minimize the decision logic.