Blind estimation of compartmental model parameters

Computation of physiologically relevant kinetic parameters from dynamic PET or SPECT imaging requires knowledge of the blood input function. This work is concerned with developing methods to accurately estimate these kinetic parameters blindly; that is, without use of a directly measured blood input function. Instead, only measurements of the output functions-the tissue ti me-activity curves-are used. The blind estimation method employed here mini mizes a set of cross-relation equations, from which the blood term has been factored out, to determine compartmental model parameters. The method was tested with simulated data appropriate for dynamic SPECT cardiac perfusion imaging with Tc-99m-teboroxime and for dynamic PET cerebral blood flow imag ing with O-15 water. The simulations did not model the tomographic process. Noise levels typical of the respective modalities were employed. From thre e to eight different regions were simulated, each with different time-activ ity curves. The time-activity curve (24 or 70 time points) for each region was simulated with a compartment model. The simulation used a biexponential blood input function and washin rates between 0.2 and 1.3 min(-1) and wash out rates between 0.2 and 1.0 min(-1). The system of equations was solved n umerically and included constraints to bound the range of possible solution s. From the cardiac simulations, washin was determined to within a scale fa ctor of the true washin parameters with less than 6% bias and 12% variabili ty. Tc-99m-teboroxime washout results had less than 5% bias, but variabilit y ranged from 14% to 43%. The cerebral blood Bow washin parameters were det ermined with less than 5% bias and 4% variability. The washout parameters w ere determined with less than 4% bias, but had 15-30% variability. Since wa shin is often the parameter of most use in clinical studies, the blind esti mation approach may eliminate the current necessity of measuring the input function when performing certain dynamic studies.