The goal of this project is to develop, and refine through testing, an accurate three-dimensional (3D) edge detection system for volume and activity quantitation in SPECT imaging. Such an edge detection system would serve as a key component when employing SPECT to quantitate activity for use in dosimetry and assessment of physiological function, as well as providing the clinical indices of lesion and organ volume. A 3D approach is required for the most accurate delineation of 3D edges in SPECT data. The proposed edge detection system is based upon calculation of the 3D gradient of the count density. A 3D search, in the direction of the gradient, is used to determine its local maximum as the edge. Preliminary results have demonstrated the success of this approach. Proposed refinements to be investigated include a multi-resolution implementation to improve the fidelity of edge detection in high noise and low contrast regions. A simulation of SPECT imaging employing the serial model of the system transfer function and analytical projection will be developed to provide testing of performance in terms of both volume determination and edge surface localization. Variations in object size, shape, closeness to other objects, inclusion of other objects, background level and texture, system transfer function, and Poisson noise level will be studied. Due to the blurring of activity by the limited spatial resolution of SPECT imaging systems, it is likely that the region determined for volume quantitation may have to be altered for optimal quantitation of activity. The software simulation of SPECT imaging will be used to test the need for such an alteration and the success of a proposed method of altering the region definition from that used for volume quantitation. Images of phantoms will be used to complete the testing of activity quantitation.

Agency
National Institute of Health (NIH)
Institute
National Cancer Institute (NCI)
Type
Research Project (R01)
Project #
1R01CA050641-01A1
Application #
3195272
Study Section
Diagnostic Radiology Study Section (RNM)
Project Start
1990-06-01
Project End
1993-05-31
Budget Start
1990-06-01
Budget End
1991-05-31
Support Year
1
Fiscal Year
1990
Total Cost
Indirect Cost
Name
University of Massachusetts Medical School Worcester
Department
Type
Schools of Medicine
DUNS #
660735098
City
Worcester
State
MA
Country
United States
Zip Code
01655