Sparse-view reconstruction and noise reduction in 99mTc-prostate specific membrane antigen prostate imaging
Year: 2025
Type: article
Abstract: PURPOSE: This preliminary study evaluates the feasibility and clinical value of sparse-view acquisition in 99m Tc-prostate specific membrane antigen (PSMA) single photon emission computed tomography (SPECT) prostate imaging. This study also compares the performance of traditional Gaussian filtering vs. an edge-preserving nonlocal means (NLM) filter in the sparse-view SPECT approach. METHODS: Ten patients with biopsy-proven prostate cancer (Gleason 4-10) underwent full-angle acquisitions (60 views), which were decimated to 30, 15, and 10 views. Images were reconstructed with ordered-subsets expectation maximization (9 iterations, 5 subsets) and postfiltered with a Gaussian kernel or NLM. Quantitative performance metrics included mean absolute percentage error (MAPE), normalized root mean square error (NRMSE), peak signal-to-noise ratio (PSNR), and normalized bias (NB). Two nuclear medicine physicians (10-12 years' experience) provided blinded visual assessments. RESULTS: Halving the views to 30 preserved diagnostic accuracy (MAPE_ROI1 ≤ 7%). Reducing to 15 views introduced sampling artifacts consistent with violation of the angular Nyquist limit, and 10 views produced unacceptable artifacts (MAPE_ROI1 > 50%). For 30-view reconstructions, NLM outperformed Gaussian, improving NRMSE by up to 39.3%, PSNR by 4.5%, and NB by 18.55%. Expert readers confirmed these trends, with high interobserver agreement (intraclass correlation coefficient = 0.83) for 30 views. CONCLUSION: Sparse-view 99m Tc-PSMA SPECT with 30 views appears feasible for routine and emergency prostate imaging in this setting. Reconstructions from 15 views showed aliasing from angular undersampling and are not recommended without anti-aliasing strategies and further validation. This work characterizes the added value of NLM postreconstruction filtering in PSMA-SPECT and motivates larger studies.
Source: Nuclear Medicine Communications
Language: English
Subfield: Radiology, Nuclear Medicine and Imaging
Field: Medicine
Domain: Health Sciences
Open Access status: hybrid
OpenAlex ID: W4416292582