Abstract

Background and Objective
The Spline Reconstruction Technique (SRT) is a fast algorithm based on a novel numerical approach to the inverse Radon transform. This study compares SRT with Filtered Back-Projection (FBP), Ordered Subset Expectation Maximization 2D (2D-OSEM), and the Tera-Tomo 3D algorithm, using phantom data at various acquisition durations and small-animal data from the Mediso nanoScan® PET/CT scanner.

Methods
The "NEMA NU 4–2008 standards" protocol was followed at five different acquisition durations. In addition to the image quality metrics outlined in the NEMA protocol, Cold Region Contrast (CRC) was assessed as an additional figure-of-merit. CRC was specifically measured in the myocardial infarction region of six male Wistar rats, with volumetric defect quantification performed using dedicated software. Recovery Coefficient and Spill-Over Ratio plots were generated as a function of Percentage Standard Deviation after applying four different Gaussian filters to the phantom reconstructions. Statistical significance was assessed using the Kruskal-Wallis test or One-way ANOVA, depending on data distribution.

Results with nanoScan PET/CT

Imaging System
This study used the Mediso nanoScan® PC (PET8/2) PET/CT scanner, a preclinical imaging system with two rings, each featuring eight detector modules. The scanner has an inner diameter of 12.6 cm, with LYSO crystal needles (1.51 × 1.51 × 10 mm³) and an axial field of view (FOV) of 98.6 mm, and a transaxial FOV of 80 mm. Coincidence data were acquired for imaging purposes.

Positron Emission Tomography (PET) is a powerful molecular imaging technique widely used in oncology, neurology, and cardiology, offering both functional and anatomical insights when combined with Computed Tomography (CT) or Magnetic Resonance Imaging (MRI). Small-animal PET/CT is particularly valuable in preclinical research and drug development. Image reconstruction is critical for PET imaging, with two main approaches: analytic methods like Filtered Back-Projection (FBP) and iterative methods like Ordered Subset Expectation Maximization (OSEM). While FBP is fast and simple, iterative methods, including OSEM, typically offer better image quality and resolution, particularly in dynamic and static PET imaging. However, iterative methods may suffer from ambiguities and biases in quantification, especially with low-count data. The Spline Reconstruction Technique (SRT) is a newer analytic method that uses a numerical approximation of the inverse Radon transform, offering potential advantages over both FBP and OSEM. SRT is based on an analytic formulation in physical space and aims to reduce streak artifacts outside the object boundary, making it well-suited for PET, CT, and SPECT applications. The Mediso nanoScan® PET/CT system, a small-animal scanner, provides high-quality imaging for preclinical studies, allowing for detailed reconstructions at various acquisition durations. This study compares SRT, FBP, and OSEM (2D and 3D) using data from the Mediso nanoScan® PET/CT system, evaluating image quality, Cold Region Contrast, and myocardial infarction quantification. The goal is to assess the performance of these reconstruction algorithms in small-animal cardiac PET studies and determine which method offers the best trade-off between accuracy and image quality.
This study found that, although SRT showed slightly higher noise in reconstructed images, it provided superior Recovery Coefficient values for small hot regions of interest compared to FBP and 2D-OSEM across all acquisition durations. At shorter acquisition times, SRT outperformed the other 2D algorithms in terms of Recovery Coefficient for all hot regions. Both phantom and animal studies demonstrated that SRT also yielded better Spill-Over Ratios and Cold Region Contrast than 2D-OSEM, even when noise levels were matched. The Tera-Tomo 3D algorithm (4 subsets, ≥13 iterations) performed best overall for all metrics. No significant differences were found between algorithms for myocardial defect size.

Conclusions. They presented a comparison of SRT, FBP, and OSEM with preclinical PET data, using the STIR library for reconstructions. The "NEMA NU 4–2008 standards" protocol, along with the Ccold metric, was applied to evaluate performance. Overall, SRT showed slightly higher noise in reconstructed images but performed well in image quality metrics, potential for accurately quantifying small hot and cold regions, especially in low-count imaging scenarios.

Full article on Sciencedirect