Presented at the 47th Annual Scientific Session of the American College of Cardiology
Atlanta, Georgia, March 29-April 1, 1998.
Note (July 2006): A newer version of this tool is the SurfaceEditor application in BioImageSuite.
|
For estimating quantitative regional myocardial surface deformation and volume changes from Cine-3D Cardiac MR images, a fast and accurate segmentation method is needed as this is a time consuming process. This handout describes a software package SurfaceEdit designed to reduce the time needed to accurately segment cardiac images.
SurfaceEdit has an intuitive user interface and can simultaneously display orthographic views of the 3d-image and multiple surface sections as well as multiple 3D surface rendering from any angle. All of the above can also be displayed in cine-mode. The colors and transparency of the surfaces can be edited to allow the user to display one surface inside another. The software development was done in C++ using the Open Inventor 3D Graphics Toolkit and the Motif toolkit on the Silicon Graphics(SGI) Platform. The machine we use at Yale to run this program is a low-end SGI O2 workstation, costing less than $6000.
The Editor Module.
We use a deformable contour-commonly known as snake- based segmentation to extract contours on short-axis slices and then form the surfaces using either a Delaunay Triangulation(accurate) or B-spline interpolation(fast). The contours are parameterized using B-Splines which allows for easy editing by moving control points, so that the expert user can easily correct for cases where the image data is not ideal. The editor also allows the user to edit upto four contours at any given time.
A Long-Axis View.
The user has almost instantaneous feedback in 3D of any changes made in the 2D contour editor. Contours can be propagated both spatially and temporally which reduces the amount of manual input necessary and takes advantage of the smooth variation of the contours across time and space.
We are currently using this program in two projects; the evaluation of Transmyocardial Laser Revascularization and as part of ongoing research in developing algorithms for the estimation of myocardial deformation from image data. In the future we are hoping to add facilities for editing data acquired in non-parallel slices, such as echocardiography.
Conclusion: This software package is designed to allow for fast, accurate segmentation with easy user feedback. This coupled with the simultaneous display of 3D images and surfaces in both static and cine-mode has resulted in a 70% reduction of the time taken to process a complete set of images(from around 16 man-hours per study). The reduction in processing time is an important step towards clinical applications of cardiac MR images.
For more information about obtaining and using SurfaceEdit please contact the first author.