The purpose of this project is collaborative development and application of advanced image processing and information processing algorithms for use in basic and clinical research. Examples of application areas include: MRI, CT, Ultrasound, PET, SPECT, confocal and electron microscopy, gel electrophoresis, and information content of macromolecular sequences. One area of emphasis has been automated determination of atheromatous plaque volume in coronary arteries through analysis of coronary ultrasound images. Although clinical decisions are commonly based upon luminal narrowing in coronary angiograms, sites of coronary artery stenosis are associated with only about half of myocardial infarctions. Newer concepts linking arterial wall changes to infarction have also indicated the need for more effective detection and quantification of arterial lesions. Intracoronary ultrasound has proven to be an excellent candidate to overcome the limitations of 2D planar image projection obtained by angiography. In this collaboration Drs. Patrick Brigger and J. Hoeg investigated two methods of image segmentation techniques: 1) transformation of the data in a more suitable, polar coordinate system, with automatic contour tracking by dynamic programming, and 2) polar coordinate transformation, followed by an active contour model for edge detection. Both systems have given good results, but the latter proved more robust with respect to imaging artifacts. (This is a continuation of Intramural Research Project Z01-RR-10495-01 BEI.) A collaboration among Drs. P. Brigger, V. Dilsizian, S. Bacharach, and G. Srinivasan sought to develop automatic methods for segmentation and quantification of gated Thallium-SPECT left ventricular myocardial perfusion images in order to assess ejection fraction (EF) and absolute volume. They have applied various image processing techniques based on dynamic programming and B-spline contour detection, as well as an elliptical coordinate transformation, matched filters to enhance the relevant image boundaries, and pre- and post-processing imaging algorithms. The segmentation schemes are extremely robust in noisy environments and have worked successfully on several image modalities. In addition, a novel scheme has been designed for computation of ejection fraction from gated perfusion images, based on the epi- and endo-cardial boundaries rather than on the endo-cardial boundary alone. The algorithms have been validated against manual border tracing on CT , MRI , PET, and Thallium- SPECT images, as well as on simulated Thallium-SPECT images where the underlying borders are known. The results have demonstrated improved accuracy compared to manual tracings and lower variances on multiple runs over the same data. (This is a continuation of Intramural Research Project Z01-RR-10494-01 BEI.) A collaboration between Drs. C-N Chen and K. Wang is ongoing to create 3D stacks of confocal microscopy images from fluorescence- tagged tissues and to visualize them to reveal structural detail. Initial studies on test images have used the IDL image manipulating system. The group is developing data handling and deconvolution algorithms for use with experimentally-obtained point-spread functions. Dr. H. Eden has begun a collaboration with Dr. R. Summers on applications of 3DVIEWNIX software (written by the Medical Image Processing Group, University of Pennsylvania) for segmentation of arteries in magnetic resonance arteriograms.
