Evaluation of analytical methods for fast and accurate image reconstruction in 3D SPECT

Abstract

The authors implemented for the first time analytical methods for image reconstruction in 3D SPECT with uniform attenuation and distance-dependent blurring. The implementation is computationally fast and requires less than 4 minutes to reconstruct a 128/spl times/128/spl times/128 image on a personal computer. The authors conducted quantitative evaluation of the reconstruction methods and revealed that, when the deblurring is considered, these reconstruction methods are highly susceptible to noise. They then developed a Wiener apodizing window function for suppression of the amplification of data noise in the reconstructed images. They also evaluated the performance of the proposed Wiener window function and the frequently used truncation window function with both noiseless and noisy data. The authors' numerical results demonstrated that the proposed Wiener window function generally yields 3D SPECT images from noisy data with visual quality better than and quantitative accuracy comparable to that of the truncation window function.