Left ventricular status, reflected in ejection fraction or end systolic volume, is a powerful prognostic indicator in heart disease. Quantitative analysis of these and other parameters from ventriculograms (cine xrays of the left ventricle) is infrequently performed due to the labor required for manual segmentation. None of the many methods developed for automated segmentation has achieved clinical acceptance. We present a method for semi-automatic segmentation of ventriculograms based on a very accurate two-stage boosted decision-tree pixel classifier. The classifier determines which pixels are inside the ventricle at key ED (end-diastole) and ES (end-systole) frames. The test misclassification rate is about 1%. The classifier is semi-automatic, requiring a user to select 3 points in each frame: the endpoints of the aortic valve and the apex. The first classifier stage is 2 boosted decision-trees, trained using features such as gray-level statistics (e.g. median brightness) and image geometry (e.g. coordinates relative to user supplied 3 points). Second stage classifiers are trained using the same features as the first, plus the output of the first stage. Border pixels are determined from the segmented images using dilation and erosion. A curve is then fit to the border pixels, minimizing a penalty function that trades off fidelity to the border pixels with smoothness. ED and ES volumes, and ejection fraction are estimated from border curves using standard area-length formulas. On independent test data, the differences between automatic and manual volumes (and ejection fractions) are similar in size to the differences between two human observers.
KEYWORDS: 3D metrology, 3D scanning, Image registration, 3D image processing, Laser scanners, 3D modeling, Reconstruction algorithms, Scanners, Laser range finders, Heart
In this paper, we present a method for simultaneous registration and surface fitting, using multiple sets of 3d points. Our motivating application is the automatic construction of CAD models from 3d surface measurements of physical objects. A wide variety of technologies are currently used to measure 3d shape [1] , and our method is applicable to the data collected by many of them. However, to simplify the discussion, we will describe the problem in terms of an idealized laser range scanner. This idealized laser range scanner captures the shape of a physical object as a set of range images. Each range image is a sample of 3d locations from those parts of the surface of the object which are visible from a single scanner viewpoint. Although single range images are adequate for some applications, in general, to model the entire surface, it is necessary to collect range images from multiple viewpoints. At some point in the processing, each range image is a set of 3d points measured relative to the corresponding scanner viewpoint. To get consistent 3d data for the entire surface of the object, the range images then need to be regisiered into a common 3d coordinate system.
Reconstruction of shapes from partial information is a problem arising in many scientific and engineering applications. We present a method for reconstructing a two-dimensional manifold from an unstructured collection of sampled points. The algorithm consists of two major steps. In the first step, we estimate the topological type of the manifold and also obtain a crude estimate of its geometry. In the second step, we improve the fit of the estimate to the data points, while keeping the topological type fixed.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.