Quantitative reconstruction of I-123 FP-CIT brain SPECT with partial volume compensation

Objectives I-123 FP-CIT brain SPECT has been used to confirm the diagnosis of Parkinson’s diseases. Semi-quantitative analysis of the striatal binding potential (BP) from the images allows monitoring of disease progress. Commonly used reconstruction methods do not include compensation for many image degrading factors, potentially resulting in large errors and variations. In this work, we evaluated different iterative reconstruction based compensations for I-123 brain SPECT imaging with partial volume compensation (PVC).

Methods A Radiology Support Devices striatal brain phantom was used. Each striatal compartment was filled with a different I-123 activity. The data were acquired on a Siemens Symbia SPECT-CT system with a typical brain protocol. A CT scan was acquired to provide attenuation maps and volume-of-interest maps (VOI). The images were reconstructed using the iterative OS-EM algorithm with different reconstruction based compensation for various degrading factors, including attenuation, detector response, collimator septal penetration and scatter, object scatter, and downscatter of I-123 high-energy photons. The previously developed perturbation-based geometric transfer matrix (pGTM) PVC method was applied to each image. The pGTM matrix was computed by adding perturbations to the measured projection data and then reconstructing. The perturbations were generated by analytically projecting the VOI maps with models of the degrading factors.

Results There was improved contrast and resolution of images obtained with compensations for more degrading factors. The underestimation of BP values was also reduced with more compensation. PVC further reduced the errors. When the pGTM PVC modeled all factors in the projection process, there was additional improvement in accuracy. The residual errors are likely due to segmentation and registration errors.

Conclusions Compensation for all image-degrading factors can improve the accuracy of BP estimates in quantitative I-123 brain SPECT. The best results in phantoms were obtained for images reconstructed with compensations for all degrading factors, including PVC.