Evaluation of an analytical image reconstruction algorithm in dynamic small animal positron emission tomography
Loading...
Date
2025-06-18
Authors
Βραχλιώτης, Αλέξανδρος
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Extended Abstract
Positron Emission Tomography (PET) is a quantitative, molecular imaging modality with proven value for diagnosis and treatment monitoring. This noninvasive imaging technique reveals functional and biological processes relevant to a wide range of fields, such as oncology, neurology, and cardiology. Small animal PET refers to imaging studies conducted on laboratory animals, typically mice and rats, using dedicated PET scanners. These animal models play a critical role in studying drug biodistribution, evaluating new imaging agents, and tracking disease progression.
PET image reconstruction has evolved into a well-established and extensively researched field. It is a fundamental process that converts raw projection data, acquired through detector measurements, into cross-sectional images. Over the years, various reconstruction methods have been developed, evolving from: (i) analytic methods based on direct mathematical models, to (ii) more advanced iterative methods, and more recently, (iii) artificial intelligence-driven methods. The variety of methods for PET reconstruction arises from the fact that there is no such thing as an optimal image reconstruction algorithm.
In this context, the current doctorate thesis aims to evaluate the performance of the Spline Reconstruction Technique (SRT) for both static and dynamic small animal PET imaging. SRT is a recently proposed algorithm based on a novel numerical implementation of an analytic representation of the inverse Radon transform. The algorithm is formulated in the “physical space” and can be applied to various imaging modalities, including PET, Computed Tomography (CT), and Single-photon Emission Computed Tomography (SPECT). In previous PET studies, SRT was primarily assessed using simulated phantom data from a clinical PET system, along with real phantom data acquired from clinical PET scanners under static imaging conditions. To achieve this goal, two comparative studies were designed to evaluate SRT against established reconstruction methods.
The purpose of the static imaging evaluation was to provide a comparison between SRT, Filtered-backprojection (FBP), Ordered Subset Expectation Maximization 2D (2D-OSEM), and the Tera-Tomo 3D algorithm, using both phantom and small animal data obtained from the Mediso nanoScan® PET/CT scanner. All 2D reconstructions were performed via software for tomographic image reconstruction (STIR open-source software, version 3.0). For the phantom study, the “NEMA NU 4-2008 small animal PET standard” was employed at five different realizations (R=5). In addition to the suggested acquisition duration of 1200 seconds, four supplementary acquisitions (12 s, 60 s, 300 s, and 600 s) were employed to investigate the performance of the algorithms under low-count imaging conditions. According to the “NEMA NU 4-2008 standards” protocol, the three main regions of the Image Quality (IQ) phantom are associated with the assessment of three image quality metrics: Percentage Standard Deviation (%STD), Recovery Coefficient (RC), and Spill-Over Ratio (SOR). %STD is a measure of image uniformity (image noise), RC quantifies the “activity recovery” in environments with zero activity background, therefore is indicative of the spatial resolution of the imaging system, and SOR is a measure of the scanner’s ability to perform scatter correction. In addition to the image quality metrics described by the “NEMA NU 4-2008 standards” protocol, Cold Region Contrast (Ccold) was also considered as a figure-of-merit. Both SOR and Ccold are two key indicators of an algorithm’s ability to distinguish between active (hot) and inactive (cold) regions. Lastly, to compare the algorithms at equal noise levels, plots of RC and SOR as a function of %STD were generated after smoothing the phantom reconstructions using four different Gaussian filters (0.5x, 1x, 1.5x, and 2x sigma). For the animal study, Ccold was adapted in the myocardial infarction region of six male Wistar rats (n=6). Furthermore, volumetric defect quantification was assessed using dedicated computer software (Munich Heart software packages). Statistical significance was determined by employing the Kruskal-Wallis test or One-way Analysis of Variance depending on the normality of the variable’s distribution.
The purpose of the dynamic imaging evaluation was to provide a comparison between SRT, FBP, and the Tera-Tomo 3D algorithm in terms of their effect on quantifying the myocardial metabolic rate of glucose (MRGlu), using small animal data obtained from the Mediso nanoScan® PET/CT scanner. For this purpose, Patlak graphical kinetic analysis was employed to noninvasively quantify MRGlu in seven male C57BL/6 mice (n=7). All analytic reconstructions were performed via software for tomographic image reconstruction (STIR open-source software, version 3.0). Analysis of all PET reconstructed images was conducted via PMOD software (version 3.506) using the inferior vena cava as the image-derived input function. Statistical significance was determined by employing the one-way analysis of variance test.
The static imaging evaluation revealed that, at the expense of slightly increased noise (%STD) in the reconstructed images, SRT resulted in higher RC values for small hot regions of interest, when compared to FBP and 2D-OSEM across all acquisition durations. Furthermore, SRT reconstructed images exhibited higher RC values, for all hot regions of interest, when compared to the other 2D algorithms at shorter acquisition durations. In both phantom and animal studies, SRT demonstrated a significant improvement over 2D-OSEM in terms of SOR and Ccold. Importantly, the advantages of SRT were maintained even after comparing the algorithms at equal noise levels. The Tera-Tomo 3D algorithm (4 subsets, iterations ≥ 13) performed significantly better compared to the other algorithms for all figures-of-merit. The absence of statistically significant differences (p>0.05) regarding myocardial defect size among the algorithms investigated suggests that SRT is a promising reconstruction method for evaluating myocardial viability. Overall, SRT appears that could be useful for the quantification of small hot regions of interest, cold regions of interest, as well as in low-count imaging applications.
The dynamic imaging evaluation revealed that the differences between the values of MRGlu obtained via SRT versus FBP, and the variants of the Tera-Tomo 3D algorithm were not statistically significant (p>0.05). Overall, SRT appears to perform similarly to the other algorithms investigated, providing a valid alternative analytic method for preclinical dynamic PET studies.
The findings of this comprehensive evaluation highlight the potential of SRT in both static and dynamic preclinical PET imaging settings.
Description
Keywords
Spline reconstruction technique (SRT), Positron emission tomography (PET), Image reconstruction, Small-animal imaging, Dynamic small-animal PET imaging, NEMA NU 4-2008 standards protocol, Quantification of the myocardial metabolic rate of glucose (MRGlu)