[Validation of noise reduction in iterative reconstruction: a simulation phantom study].

Area of Science:

• Medical Imaging
• Nuclear Medicine
• Image Processing

Background:

• Iterative reconstruction in SPECT is sensitive to noise, which increases with lower photon counts and more iterations.
• Image filtering is crucial for SPECT reconstruction quality but often applied subjectively.
• Evaluating filtering strategies is essential for optimizing SPECT image analysis.

Purpose of the Study:

• To assess the impact of pre-reconstruction versus post-reconstruction filtering on iterative SPECT image reconstruction.
• To compare the effectiveness of Butterworth (BW) and Gaussian (GA) filters applied at different stages.

Main Methods:

• Simulation phantoms were used to generate projection data.
• Ordered Subsets Expectation Maximization (OSEM) was employed for reconstruction, with and without resolution recovery.
• Pre- and post-reconstruction filtering using BW and GA filters with varying parameters were evaluated.

Main Results:

• Pre-reconstruction filtering yielded lower Full Width at Half Maximum (FWHM) (5.1 mm) compared to post-reconstruction (6.7 mm).
• Coefficient of Variation (CV) was significantly lower with pre-reconstruction filtering (7.5%) versus post-reconstruction (11.6%), indicating better noise reduction.
• Normalized Mean Squared Error (NMSE) was also lower with pre-reconstruction filtering, suggesting improved overall image accuracy.

Conclusions:

• Pre-reconstruction filtering, particularly with Butterworth filters, achieves SPECT image quality comparable to post-reconstruction filtering.
• Applying pre-reconstruction filtering is a beneficial strategy for iterative reconstruction methods, enhancing image quality and noise control.