Development of Software Tools for Evaluation of Crosstalk Compensation Method for Simultaneous Dual Isotope Myocardial SPECT Imaging



Xiaolan Wang, Martin Stumpf, and Eric C. Frey



Myocardial perfusion SPECT (MPS) studies are the most common clinical SPECT studies. A common protocol for performing MPS uses a dual isotope protocol which uses Tl-201 chloride to obtain the rest image and a Tc-99m labeled agent to obtain the stress image. Due to crosstalk between the images for the two isotopes, the current clinical standard is to acquire the Tl-201 and Tc-99m images sequentially. However, simultaneous acquisition would have the advantage of reduced acquisition time and provide images that are perfectly registered in space and time. To facilitate this, we have previously developed model-based methods for crosstalk compensation which provide a substantial improvement in image quality. We are in the process of evaluating this method using a set of > 100 clinical dual isotope studies. The comparison will be performed by presenting physicians with a set of short, horizontal long, and vertical long axis slices of images reconstructed from separate acquisition and the simultaneously acquired Tl-201 images with and without crosstalk compensation. We currently have software tools for reorienting these images either as command-line programs or interactive programs on the GE Xeleris workstation. The former are difficult to use while the latter require manually reorienting each image, even those reconstructed from the same projection data. The purpose of this project is to develop a software tool to interactive reorienting of the slices such that the same reorientation parameters can be applied to all images obtained from the same set of projection data.



To reduce development time, we developed the program in MATLAB. The program uses a wizard-like paradigm. The user reads in an image and feeds in the cutoff frequency and filter order for the smoothing Butterworth filter before reorientation.  

The reorientation process starts with drawing on a transaxial view slice a "rubber line" that passes through the heart, dividing the chamber into approximately equal parts. Then a click of "Apply" button is all needed to generate the vertical long axis slices via two subroutines: first, rotate every slice of the image as a 2D stack so that the pixels on the rubber line lie on the x-axis; second   the 3D image is restacked in such way that   slices perpendicular to the y-axis in the rotated image are saved as slices perpendicular to the z-axis in the output image.

The horizontal long and short axis slices can be obtained in a similar manner. When shown the vertical long axis slices, the user draws a line that separates the chamber into equal halves. Click "Apply" again, and the rotation and restacking of the slices will produce the desired output image. Furthermore, a parameter file containing all filter and point coordinates information involved in this process will be generated which can be applied with the command-line tool, "cardproc"

Note that the program provides a separate display window for a "preview" display of the output of current operation. So before moving on to the next step, the user can play around with any of the filter parameters as well as  the rubber lines positions until satisfactory results are reached.



The program has been written and the results compared to output from the previous command-line tool, "cardproc". The resulting horizontal long axis slices differ by no more than 2% at the most inconsistent pixel, and the resulting horizontal long and short axis slices only show a less than 1% difference at a couple of  peak and background pixel values.  In all, this program obtains the almost equivalent level of accuracy as "cardproc", the command-line tool currently in wide applications.



We have developed a tool for performing reorienting myocardial perfusion SPECT images. In the future this tool will be used as part of a clinical evaluation of a crosstalk compensation method for myocardial perfusion imaging.



This work was funded by Public Health Service Grant R01 EB 00288.