Quantitative brain SPECT (single-photon emission computed tomography) requires simultaneous compensation for photon attenuation, scatter, and depth-dependent detector-resolution variation, as well as suppression of Poisson noise. Conventional filtered backprojection (FBP) and iterative (ITT) reconstructions either lack quantitative accuracy or demand intensive computation. The long-term objective of this project is to develop accurate analytical (noniterative) reconstruction techniques for object-specific brain SPECT. The analytical techniques will achieve the quantitative accuracy of ITT, while maintaining the computational efficiency of FBP.
The specific aims of this proposal are: (1) To develop the analytical techniques-treat Poisson noise using Anscombe transform and Wiener filter; remove scatter contribution by multiple energy-window acquisitions; restore detector resolution via depth- dependent deconvolution; and compensate for head attenuation without transmission scans by extending the head boundary appropriately (where the head boundary is determined from scatter-data reconstructions) and inverting attenuated Radon transform accurately via Fourier transforms: (2) To evaluate reconstruction accuracy by anthropomorphic brain-phantom studies - use the Hoffman brain phantom to simulate emission tracer distribution; add a bone-equivalent shell and a plastic layer to the phantom for the skull and scalp; reconstruct the attenuation map from transmission scans of a three-head SPECT system with an external line source; acquire emission data by the SPECT system; construct the SPECT resolution kernel from point-source measurements at various depths; reconstruct the emission data analytically using both the object- specific and the enlarged-uniform attenuation maps; and compare the reconstructions using region-of-interest (ROI) studies by several criteria. (3) To implement the analytical techniques for clinical studies - Ten sets of patient transmission and emission data will be acquired using the SPECT system. The reconstructions by the analytical techniques will be compared to those obtained by FBP and ITT Bayesian methods. Clinical benefits will be drawn from quantitative measures of selected ROIs of the reconstructions. These accurate and efficient unified analytical techniques should have a significant impact on our ability to diagnose various brain diseases and to probe the metabolism.