Improved quantitation of dynamic SPECT via fully 4-D joint estimation of compartmental models and blood input function directly from projections

Quantitative kinetic analysis of dynamic cardiac single-photon emission computed tomography (SPECT) data provides unique information that can enable improved discrimination between healthy and diseased tissue, compared to static imaging. In particular, compartmental model analysis can provide quantitative measures of myocardial perfusion, viability, and coronary flow reserve. In this work we investigate whether precision of kinetic parameter estimates is improved by additional temporal regularization provided by estimating compartmental models directly from projection data, rather than using "semidirect" methods that estimate time-activity curves first and then fit compartmental models to the curves. Methods are implemented to accelerate fully 4-D direct joint estimation of compartmental models for tissue volumes and B-spline time-activity curves for the blood input and other volumes that do not obey a compartmental model. Computer simulations of a dynamic ^99mTc-teboroxime cardiac SPECT study show that the additional temporal regularization provided by direct compartmental modeling results in improved precision of parameter estimates, as well as comparable or improved accuracy. Notably, for small myocardial defects the sample standard deviation of uptake and washout parameters was reduced by 17-41%. These results suggest that direct joint estimation of compartmental models and blood input function can improve quantitation of dynamic SPECT. These methods can also be applied to dynamic positron emission tomography (PET).