Quantitative study of iterative reconstruction algorithms of spect/ct in bone scan: a clinical and phantom study

• Main Author: Hao, Lau Lik
• Format: Monograph
• Language: English
• Published: Universiti Sains Malaysia 2024
• Subjects: R Medicine; RC71-78.7 Examination. Diagnosis Including raiography; RD701-811 Orthopedic surgery
• Online Access: http://eprints.usm.my/61352/1/Lau%20Lik%20Hao-E.pdf; http://eprints.usm.my/61352/

The integration of single photon emission computed tomography (SPECT) with computed tomography (CT), along with advancements in iterative image reconstruction algorithms, significantly enhances the feasibility of SPECT quantification in bone scan. Quantitative bone SPECT enables the precise measurement of radiotracer accumulation in bone lesions. This capability allows for the accurate assessments of the presence and extent of bone abnormalities, thereby improving diagnostic accuracy of bone scan. This study evaluates the impact of iterative reconstruction algorithms with various attenuation correction methods on SPECT quantification accuracy and image quality in bone scan across different iteration numbers in both phantom and clinical settings. In the phantom study, spheres in the NEMA 2012/ IEC 2008 phantom were filled with 300 kBq/ml of mixture of K2HPO4 solution and a 99mTc source, while the background region contained only 30 kBq/ml of 99mTc source, establishing a tumor-to-background ratio (TBR) of 10:1. The phantom underwent bone imaging using the standard protocol applied at Hospital Universiti Sains Malysia (HUSM). In the clinical study, a pelvic bone scan image with multiple lesions was retrieved from the XelerisTM workstation. Both phantom and clinical images were reconstructed using MLEM-CHANG, OSEM-CHANG, and OSEM-CT, with varying iteration products (4, 8, 12, 16, and 20 iterations for MLEM; 40, 80, 120, 160, and 200 iterations for OSEM). Quantitative analysis of activity concentration, recovery coefficient (RC), standardized uptake value (SUV), signal-to-noise ratio (SNR), and noise were performed using Dosimetry Toolkit and Q.Metrix software. OSEM-CT (-73.3% to 6.7%) demonstrated the smallest percentage difference between measured and actual activity concentration (300 kBq/ml) across all sphere volumes and iteration numbers compared to MLEM-CHANG (-86.7% to -43.3%) and OSEM-CHANG (-83.3% to -33.3%). For all algorithms, increasing the iteration numbers elevated RC, SUV, and noise, while SNR dropped. In the phantom study, there were no significant difference in RC and SNR among the algorithm pairs (MLEM-CHANG vs OSEM-CHANG, MLEM-CHANG vs OSEM-CT, and OSEM-CHANG vs OSEM-CT) across different iteration numbers (p>0.05), as tested using the Kruskal-Wallis test with post-hoc Bonferroni’s correction. In the clinical study, significant differences in SUV were displayed between MLEM-CHANG vs OSEM-CT and OSEM-CHANG vs OSEM-CT at all iteration numbers (p< 0.05). Additionally, the SNR of the lesions in clinical study showed significant differences between MLEM-CHANG and OSEM-CT at iterations of 8, 12 and 16 (p< 0.05). In summary, OSEM-CT illustrated higher activity concentration accuracy, RC, SUV, and SNR, along with lower noise level compared to OSEM-CHANG and MLEM-CHANG. Thus, OSEM-CT is recommended for accurate SPECT quantification and optimal image quality in bone scan.