AutoSPECT offers several automatic processing modules for single-photon emission computed tomography (SPECT) studies. Although it is mainly designed for cardiac data, many of its functionalities can be applied to other types of SPECT studies.

AutoSPECT was the first software package to provide automatic reorientation of three-dimensional, transaxial myocardial perfusion SPECT images. It was initially tested on more than 400 studies with a success ratio of 98.5%.

Each module has associated pages that present data and the controls necessary to perform the specific task for which the page has been designed.

The program can be used interactively on one or more datasets or in batch mode to process data without further intervention from the user. If matching rest and stress datasets are provided, AutoSPECT will automatically operate in dual mode.



The purpose of this first module (left)is to reconstruct the data from projections using a variety of algorithms (including iterative reconstruction). The reconstruction page shows the filter and its parameters, as well as the data. From top to bottom: raw projection data, filtered projection data and transverse reconstructed slices.


Once the data has been reconstructed, this module (right) processes transaxial images to reorient them automatically into short-axis (oblique) myocardial perfusion SPECT images.

First, this module determines the extent of the left ventricle, thus limiting subsequent operations to a restricted amount of data. This is done by segmenting the left ventricle by iterative clustering and rule-based location/size/shape criteria.

The final mid-myocardial surface is then determined by an iterative process that incorporates additional constraints of shape and texture and uses information from the nonsegmented transaxial image.



This page presents information in a different manner. Raw projections (corrected and uncorrected) are shown in the top row. The bottom part shows the sinogram and the correction control window.

Each row in the sinogram corresponds to a projection in the original raw data. The pixel values in the sinogram for a particular row are the values of the voxels located along the intersection of the chosen transverse plane and raw projection plane.

This screenshot is an example of AutoSPECT operating in dual mode with rest and stress datasets.


The filter page allows the user to view the effect of several filters on the raw data. This module can be used before reconstruction to select a more appropriate filter than the default one.

The upper and lower part of the page can be modified automatically to generate a bracket of parameters around the values of the central filter.

In this case, the same filter type with three different parameters has been selected.

Software Availability
Have Questions or Need Help?

Contact us if you have questions, or wish to learn more about artificial intelligence research at Cedars-Sinai.

8700 Beverly Blvd., Metro 203
Los Angeles, CA 90048