Multistart optimisation algorithm for joint spatial and kinetic parameter estimation in dynamic ECT

Presents a multistart optimisation algorithm for the joint estimation of spatial and kinetic parameters for dynamic emission computed tomography (ECT) studies in which tomographic projections are sampled serially in time. Parameter estimation is effected using a projected descent algorithm which refines initial parameter states generated stochastically via simulated annealing. The exponential kinetics of the compartmental model are approximated using an orthogonal basis set which reduces the dimensionality of the parameter space and addresses the problem of the non-uniqueness of exponential sums. The algorithm produces encouraging results when used to fit a five elliptical region single compartment model to the sinogram (obtained over a single camera rotation) of a dynamic phantom