Effects of scatter modeling on time-activity curves estimated directly from dynamic SPECT projections
Quantitative analysis of uptake and washout of cardiac single photon emission computed tomography (SPECT) radiopharmaceuticals has the potential to provide better contrast between healthy and diseased tissue, compared to conventional reconstruction of static images. Previously, we used B-splines to model time-activity curves (TACs) for segmented volumes of interest and developed fast least-squares algorithms to estimate spline TAC coefficients and their statistical uncertainties directly from dynamic SPECT projection data. This previous work incorporated physical effects of attenuation and depth-dependent collimator response. In the present work, we incorporate scatter and use a computer simulation to study how scatter modeling affects directly estimated TACs and subsequent estimates of compartmental model parameters. An idealized single-slice emission phantom was used to simulate a 15 min dynamic {sup 99m}Tc-teboroxime cardiac patient study in which 500,000 events containing scatter were detected from the slice. When scatter was modeled, unweighted least-squares estimates of TACs had root mean square (RMS) error that was less than 0.6% for normal left ventricular myocardium, blood pool, liver, and background tissue volumes and averaged 3% for two small myocardial defects. When scatter was not modeled, RMS error increased to average values of 16% for the four larger volumes and 35% for the small defects. Noise-to-signal ratios (NSRs) for TACs ranged between 1-18% for the larger volumes and averaged 110% for the small defects when scatter was modeled. When scatter was not modeled, NSR improved by average factors of 1.04 for the larger volumes and 1.25 for the small defects, as a result of the better-posed (though more biased) inverse problem. Weighted least-squares estimates of TACs had slightly better NSR and worse RMS error, compared to unweighted least-squares estimates. Compartmental model uptake and washout parameter estimates obtained from the TACs were less sensitive to whether or not scatter was modeled, compared to the TACs themselves.
- Research Organization:
- Lawrence Berkeley National Lab. (LBNL), Berkeley, CA (United States)
- Sponsoring Organization:
- USDOE Director. Office of Science. Office of Biological and Environmental Research. Medical Sciences Division; U.S. Department of Health and Human Services Grant R01-EB001970/R01-HL50663 (US)
- DOE Contract Number:
- AC03-76SF00098
- OSTI ID:
- 828133
- Report Number(s):
- LBNL-52702; R&D Project: 864E1C; TRN: US0403845
- Resource Relation:
- Conference: 2003 IEEE Nuclear Science Symposium and Medical Imaging Conference, Portland, OR (US), 10/19/2003--10/25/2003; Other Information: PBD: 29 Oct 2003
- Country of Publication:
- United States
- Language:
- English
Similar Records
Circumferential profiles for region-based analysis of dynamic SPECT data
Dynamic cardiac SPECT computer simulations for teboroxime kinetics