Optimal Reconstruction Filter Parameters for Multi-Headed Brain SPECT: Dependence on Count Activity

Abstract

Objective: Reconstruction parameters must be optimized to approach the improved spatial resolution that is theoretically possible from modern SPECT imaging devices. One important reconstruction consideration is proper selection of the back-projection filter parameters. The aim of this study was to determine which Butterworth filter optimized image resolution for brain SPECT scans.

Methods: Ten normal subjects underwent brain SPECT following an intravenous injection of 15–20 mCi ^99mTc-HMPAO (5 subjects) and 30–35 mCi ^99mTc-HMPAO (5 subjects). Subjects were separated into two groups based on the counts per pixel in the anterior projection image: 5 subjects with low-count images (10.4–13.9 counts/pixel) and 5 subjects with high-count images (18.3–21.5 counts/pixel). All subject projection data were reconstructed at cutoff frequencies (Fc) ranging from 0.15–0.3 Nyquist, with orders ranging from 2–12 at each frequency cutoff. Optimal image selection was determined by blinded subjective assessment by three nuclear medicine faculty capable of evaluating brain SPECT image quality.

Results: Results demonstrated that for low-count images, optimum Butterworth filter reconstruction parameters were: Fc = 0.20, order 6. For high-count images, optimum parameters were: Fc = 0.225, order 7.

Conclusion: The proper selection of Fc is the most critical parameter for optimization of image reconstruction quality.