Reconstruction of Dynamic Renal Tomographic Data Acquired by Slow Rotation

Abstract

Nuclear medicine renal studies can be performed using slow-rotation SPECT, but reconstruction of such data is largely underdetermined. Methods: A new method of reconstruction of data acquired using slow camera rotations was developed. In this method we used a factor model of the data in which the factors and factor coefficients were determined by modeling their relationship directly with the projection measurements. This was done by solving a least-squares problem that fits the projections of factors and factor coefficients to the projection data with nonnegativity constraints imposed on the solution. The method was tested on computer simulations and applied to experimental renal ^99mTc-mercaptoacetyltriglycine canine and patient studies. Results: Computer simulations showed that the extracted time–activity curves of kidneys agreed well with the simulated curves for data with noise levels similar to those in the experimental studies. In the canine study, the method showed that >2 factors were necessary to adequately reproduce the kinetics of the kidney. In the patient study, the method was able to extract separate factors that correspond to the kidney cortex and the kidney pelvis. Conclusion: The computer simulation, the canine study, and the patient study all show that reconstructions of the data obtained with 1 detector displayed artifacts, whereas reconstructions of the data obtained with 2 and 3 detectors were free of artifacts. Computer simulations showed that the method gives accurate results that allow quantitation.