Anatomically CT-guided iterative reconstruction using KEM for sparse projection 99mTc-MDP Bone SPECT/CT: quantitative...

Area of Science:

• Nuclear Medicine
• Medical Imaging
• Radiology

Background:

• Bone Single-Photon Emission Computed Tomography (SPECT)/CT is crucial for staging prostate cancer (PCa) and monitoring bone metastases.
• Traditional Ordered Subset Expectation Maximization (OSEM) is widely used for SPECT reconstruction, but can be limited by noise and acquisition time.
• Kernel Expectation Maximization (KEM) is an alternative reconstruction algorithm with potential advantages for image quality and efficiency.

Purpose of the Study:

• To systematically compare the performance of Kernel Expectation Maximization (KEM) against Ordered Subset Expectation Maximization (OSEM) for anatomically CT-guided bone SPECT reconstruction.
• To evaluate the impact of sparse-projection data on image quality and diagnostic accuracy using both KEM and OSEM.
• To determine if KEM can reduce SPECT/CT scan times without sacrificing diagnostic quality.

Main Methods:

• Sixty-seven prostate cancer patients with bone metastases underwent SPECT/CT scans.
• SPECT images were reconstructed using both KEM and OSEM with full and sparse-projection data (1/2, 1/4, 1/16).
• Tumor-to-background ratios (T/Bmax, T/Bmean) and image quality metrics (MAE, MSE, PSNR, SSIM, NRMSE) were quantified and compared.

Main Results:

• KEM reconstructions showed significantly higher tumor-to-background ratios (T/Bmax, T/Bmean) than OSEM across all projection conditions (P < 0.05).
• KEM demonstrated superior image quality metrics, including PSNR, SSIM, and NRMSE, compared to OSEM under sparse-projection conditions (all P < 0.05).
• KEM with 1/4 projection achieved image quality comparable to OSEM with 1/2 projection, and KEM with 1/16 projection outperformed OSEM with full projection in lesion conspicuity.

Conclusions:

• The Kernel Expectation Maximization (KEM) algorithm significantly enhances lesion conspicuity in bone SPECT/CT for prostate cancer.
• KEM maintains diagnostic accuracy comparable to OSEM even with significantly reduced projection data.
• KEM offers a promising approach to optimize SPECT/CT workflows by reducing acquisition time while preserving diagnostic image quality.