High-dimensional Reconstruction Algorithms In Molecular Imaging With Dynamic Cardiac SPECT

Dynamic SPECT imaging,in conjunction with mathematical model-based analysis of regional kinetics of the radiotracer,allows for the quantitation of functional parameters,such as influx rate and washout rate.However,there are some obstacles for the clinical prevalence of the dynamic SPECT technique.First of all,In SPECT systems with rotating cameras,due to the lower detection efficiency and the slow rotation of the camera,it is difficult to derive the time activity curve of the radiotracer directly from the reconstructed images.In this dissertation,a new dynamic reconstruction algorithm based on the dimensional reduction in spatial and temporal domains was proposed.The static image reconstructed from the late static projections was introduced as the prior information to reduce the dimension in spatial domain.The adaptively non-uniform B-splines functions selected were utilized to reduce the temporal dimension in temporal domain.The proposed algorithm verfied through the simulation studies as well as the rat to be able to acquire more accurate time activity curve of radiotracer with much shorter time than that from the dynamic reconstruction algorithm previously proposed.The less computational time of the proposed algorithm contributes to the implementation of the dynamic SPECT imaging in the clinical SPECT system.With the proposed dynamic algorithm,the Spontaneously Hypertensive Rats and the WKY normal rats were imaged at four time points?ages of rats:6 months,11-12 months,18 months and21-22 months?in the longitudinal evaluation of sympathetic nervous system and perfusion study using dynamic SPECT ¹²³I-meta-iodobenzylguanidine?MIBG?and ²⁰¹Thallium?Tl?imaging technique.The standard uptake value?SUV?from static imaging study and the influx rate,washout rate and the distribution volume?DV?from the dynamic imaging study were employed to evaluate the sympathetic innveration and perfusion function during the process from hypertension to the hypertrophy and finally to heart failture.The DV from the dynamic SPECT ¹²³I-MIBG imaging showed significant difference between the SHRs and WKY rats since rats were 6 months old,while the SUV from static SPECT ¹²³I-MIBG imaging had the ability since rats were 18 months old.The DV from dynamic SPECT ²⁰¹Tl imaging exhibited the significant difference between the SHRs and WKY since rats were 18 months old while the SUV from static ²⁰¹Tl imaging did not display any difference between SHRs and WKY during the whole study.These results indicated that the dynamic SPECT imaging is able to diagnose the hypertrophy or heart faulre at an earlier stage than the static SPECT imaging.The abnormal sympathetic innervation function appears much earlier than the structural abnormality.Secondly,the long acquisition time of the dynamic SPECT acquisition protocol also prevented the prevalence of dynamic SPECT imaging technique.We proposed an optimization technique to adjust the dynamic SPECT acquisition protocol to shorten the acquisition time through adjusting non-unifom acquisition intervals at each acquisition angle.The optimization technique is based on the dimensional reduction in spatial and temporal domain dynamic reconstruction algorithm.The acquisition intervals were adjusted through adjusting the knots?or control points?of B-spline basis functions.By iteratively adjusting the knots?or control points?of B-spline basis functions,new TACs were created according to two rules:accuracy of the TAC and precision of the estimated parameters.The accuracy criterion requires a low relative-entropy between the estimated left ventricular blood pool TAC and the true TAC,while the D-optimal technique forces the parameters estimated to be as precise as possible.Based on the final knots obtained,a new protocol was built and evaluated through digital simulations of dynamic cardiac ¹²³I-MIBG imaging in SHRs and WKY rats with three different kinds of kinetic parameters.Compared with the previous dynamic protocol,the optimzed protocol is able to acquire the accurate TACs with less computational time and storage space.The optimized protocol performs well with low injected radiotracer dose.Heart motion is another factor affecting the image quality.In this dissertation,we proposed a new algorithm to reconstruct more cardiac phased images in a cardiac cycle for gated-SPECT imaging without increasing the frame of the projection.These cardiac phased images were modeled as the superposition of a low-rank matrix and a sparse matrix,where the low rank matrix is corresponding to the stationary parts of the heart over the cardiac cycle and the sparse matrix is corresponding to the motion parts of heart that was represented sparsely as the product of the down-sampled B-spline basis functions and their coefficients.The simulation results indicated that the proposed reconstruction algorithm performed better than the frame-by-frame algorithm on reconstruction accuracy,uniformity of myocardium and defect detectability.Meanwhile,the proposed algorithm was able to reconstruct more cardiac phased images during one cardiac cycle without increasing injected dose,warranting further investigations to the gated-SPECT imaging without ECG in the future.