Anastasios Gaitanis

PET image reconstruction: A stopping rule for the MLEM algorithm based on properties of the updating coefficients

An empirical stopping criterion for the 2D-maximum-likelihood expectation-maximization (MLEM) iterative image reconstruction algorithm in positron emission tomography (PET) has been proposed. We have applied the MLEM algorithm on Monte Carlo generated noise-free projection data and studied the properties of the pixel updating coefficients (PUC) in the reconstructed images. Appropriate fitting lead to an analytical expression for the parameterization of the minimum value in the PUC vector for all non-zero pixels for a given number of detected counts, which can be employed as basis for the stopping criterion proposed. These results have been validated with simulated data from real PET images.

Evaluation of a spline reconstruction technique: Comparison with FBP, MLEM and OSEM

An efficient, two-dimensional, analytic, Spline Reconstruction Technique (SRT) has been presented earlier in the literature. This technique involves the Hilbert transform of the sinogram which is approximated in terms of natural cubic splines. The aim of this study is to evaluate the SRT algorithm using Monte-Carlo simulated sinograms and real PET data, in comparison with three commonly used reconstruction algorithms: FBP, MLEM and OSEM. For the simulation studies, a digital Hoffman phantom, a NEMA-like and a Derenzo phantom were employed, and Monte Carlo methods were used for the simulation of the activity distribution in the source and the resulting generation of positron-electron annihilations. No noise, scatter and absorption conditions were assumed. The phantoms were generated with different image activities. The relevant modeled system was a single-ring tomograph with 234 scintillation crystals. Image grids with an image size of 128 × 128 pixels were employed. For the studies of real data, PET sinograms of an FDG injected mouse and a NEMA and Derenzo phantom were acquired from an ARGUS-CT small animal PET/CT system. Both the simulated and real sinograms were reconstructed using the SRT algorithm and the reconstructed images were compared to those of FBP, MLEM and OSEM. The contrast and SNR were calculated for the simulated NEMA-like and Hofmann phantom by drawing ROIs within the images. Our results indicate that SRT and FBP give reconstructed images of comparable quality with respect to the number of counts. Striking artifacts become worse at lower total counts for both methods. SRT reconstructed images exhibit higher SNR in comparison with FBP and, in some cases, in comparison with MLEM and OSEM. SRT reconstructed images exhibit higher contrast over FBP but not over MLEM and OSEM. The reconstruction time for SRT was about 20 sec per slice, hence SRT is faster than MLEM and OSEM (for high activity images), but slower than FBP. In conclusion, SRT is a linear algorithm which can serve as a good alternative to FBP, providing images with higher contrast and SNR values. Furthermore, it has the crucial advantage that it can accommodate complicated detector geometries.

Intratumoral accumulation of podoplanin-expressing lymph node stromal cells promote tumor growth through elimination of CD4+ tumor-infiltrating lymphocytes

ABSTRACT The beneficial effects of checkpoint blockade in tumor immunotherapy are limited to patients with increased tumor-infiltrating lymphocytes (TILs). Delineation of the regulatory networks that orchestrate the presence of TILs holds great promise for the design of effective immunotherapies. Podoplanin/gp38 (PDPN)-expressing lymph node stromal cells (LNSCs) are present in tumor stroma; however, their effect in the regulation of TILs remains elusive. Herein we demonstrate that intratumor injection of ex-vivo-isolated PDPN+ LNSCs into melanoma-bearing mice induces elimination of TILs and promotes tumor growth. In support, PDPN+ LNSCs exert their function through direct inhibition of CD4+ T cell proliferation in a cell-to-cell contact independent fashion. Mechanistically, we demonstrate that PDPN+ LNSCs mediate T cell growth arrest and induction of apoptosis to activated CD69+CD4+ T cells. Importantly, LTbR-Ig-mediated blockade of PDPN+ LNSCs expansion and function significantly attenuates melanoma tumor growth and enhances the infiltration and proliferation of CD4+ TILs. Overall, our findings decipher a novel role of PDPN-expressing LNSCs in the elimination of CD4+ TILs and propose a new target for tumor immunotherapy.

Modeling and Simulation of 4D PET-CT and PET-MR Images

The driving force in the research and development of new hybrid PET-CT/MR imaging scanners is the production of images with optimum quality, accuracy, and resolution. However, the acquisition process is limited by several factors. Key issues are the respiratory and cardiac motion artifacts that occur during an imaging session. In this article the necessary tools for modeling and simulation of realistic high-resolution four-dimensional PET-CT and PET-MR imaging data are described. Beyond the need for four-dimensional simulations, accurate modeling of the acquisition process can be included within the reconstruction algorithms assisting in the improvement of image quality and accuracy of estimation of physiologic parameters from four-dimensional hybrid PET imaging.

Evaluation of a Spline Reconstruction Technique for SPECT: Comparison with FBP and OSEM

An efficient, two-dimensional, analytic, Spline Reconstruction Technique (SRT) for SPECT has been presented earlier in the literature. The SRT is a reconstruction algorithm based on an analytic formula of the inverse Attenuated Radon Transform which results from the approximation of the sinogram in terms of cubic splines. The algorithm incorporates the attenuation map obtained by a CT scan, to provide attenuation corrected images. The aim of this study is to evaluate the performance of the SRT algorithm, using phantoms and real SPECT/CT data, by comparing it with two commonly used reconstruction algorithms: FBP and OSEM.

The SRT reconstruction algorithm for semiquantification in PET imaging

PURPOSE The spline reconstruction technique (SRT) is a new, fast algorithm based on a novel numerical implementation of an analytic representation of the inverse Radon transform. The mathematical details of this algorithm and comparisons with filtered backprojection were presented earlier in the literature. In this study, the authors present a comparison between SRT and the ordered-subsets expectation-maximization (OSEM) algorithm for determining contrast and semiquantitative indices of (18)F-FDG uptake. METHODS The authors implemented SRT in the software for tomographic image reconstruction (stir) open-source platform and evaluated this technique using simulated and real sinograms obtained from the GE Discovery ST positron emission tomography/computer tomography scanner. All simulations and reconstructions were performed in stir. For OSEM, the authors used the clinical protocol of their scanner, namely, 21 subsets and two iterations. The authors also examined images at one, four, six, and ten iterations. For the simulation studies, the authors analyzed an image-quality phantom with cold and hot lesions. Two different versions of the phantom were employed at two different hot-sphere lesion-to-background ratios (LBRs), namely, 2:1 and 4:1. For each noiseless sinogram, 20 Poisson realizations were created at five different noise levels. In addition to making visual comparisons of the reconstructed images, the authors determined contrast and bias as a function of the background image roughness (IR). For the real-data studies, sinograms of an image-quality phantom simulating the human torso were employed. The authors determined contrast and LBR as a function of the background IR. Finally, the authors present plots of contrast as a function of IR after smoothing each reconstructed image with Gaussian filters of six different sizes. Statistical significance was determined by employing the Wilcoxon rank-sum test. RESULTS In both simulated and real studies, SRT exhibits higher contrast and lower bias than OSEM at the cold lesions. This improvement is achieved at the expense of increasing the noise in the reconstructed images. For the hot lesions, SRT exhibits a small improvement in contrast and LBR over OSEM with 21 subsets and two iterations; however, this improvement is not statistically significant. As the number of iterations increases, the performance of OSEM improves over SRT but again without statistical significance. The curves of contrast and LBR as a function of IR after Gaussian blurring indicate that the advantage of SRT in the cold regions is maintained even after decreasing the noise level by Gaussian blurring. CONCLUSIONS SRT, at the expense of slightly increased noise in the reconstructed images, reconstructs images of higher contrast and lower bias than the clinical protocol of OSEM. This improvement is particularly evident for images involving cold regions. Thus, it appears that SRT should be particularly useful for the quantification of low-count and cold regions.

Evaluation of modified median root prior on a myocardium study, using realistic PET/MR data

One way of treating the partial volume effect in PET image reconstruction is by using anatomical information from other imaging modalities (MRI or CT). The a priori information of a maximum a posteriori reconstruction algorithm is defined from the anatomical images. In this paper the ordered subsets modified median root prior one step late (OS-MMRP-OSL) algorithm [1], which uses information derived from MR images, is evaluated in a computationally simulated PET FDG myocardium study. The algorithm was implemented in STIR (Software for Tomographic Image Reconstruction) [2], (http://stir.sourceforge.net). Realistic PET data have been used, to compare the standard ordered subsets median root prior one step late (OS-MRP-OSL) algorithm with the OS-MMRP-OSL algorithm using well-aligned segmented and non-segmented MR images. In some cases the quantitative results indicate lower bias (by 6.5%) for OS-MMRP-OSL using segmented MR images and decreased root mean square error (RMSE) by 3%. Moreover, we have improvement in edge preservation.

aSRT: A new analytic reconstruction algorithm for SPECT

We present aSRT, the attenuated Spline Reconstruction Technique for single photon emission computerized tomography (SPECT): an innovative image reconstruction algorithm based on an analytic formulation of the Inverse Attenuated Radon Transform (IART). aSRT involves the calculation of the Hilbert transform of the linear attenuation coefficient and the Hilbert transform of two sinusoidal functions of the so-called attenuated sinogram. This is accomplished by utilizing the attenuation information provided by computerized tomography (CT) scans and by employing custom-made cubic splines interpolation. The purpose of this work is: (i) to present the mathematics of aSRT, (ii) to reconstruct simulated and real SPECT/CT data using aSRT, and (iii) to evaluate aSRT by comparing it to filtered back-projection (FBP) and to ordered subsets expectation minimization (OSEM). Simulation studies were performed by using an image quality (IQ) phantom and an appropriate attenuation map. Reconstructed images were generated for 45, 90 and 180 views over 360 degrees and twenty realizations of Poisson noise were created at three different noise levels, namely 100% (NL1), 50% (NL2) and 10% (NL3) of the total counts, respectively. Moreover, real attenuated SPECT sinograms were reconstructed taking into account the corresponding attenuation map provided by a clinical SPECT/CT system. Comparisons between aSRT, FBP and OSEM reconstructions were evaluated using contrast, bias and image roughness. Plots are presented as a function of image roughness. These results suggest that aSRT can efficiently produce accurate attenuation-corrected reconstructions for simulated phantoms, as well as clinical data. The algorithm provides good quality images with significant improvement over FBP and comparable to OSEM. aSRT, by incorporating the attenuation correction within itself, may provide an improved alternative to FBP.

SPNsim: A database of simulated solitary pulmonary nodule PET/CT images facilitating computer aided diagnosis

The aim of the present work was to design and develop a database of simulated solitary pulmonary nodules (SPN) in pairs of computed tomography (CT) and positron emission tomography (PET) images, using Monte Carlo (MC) simulation methods. We have developed an SPN image modeling pipeline to feed the database entitled SPNsim. The database is web-accessible and it is contains two subsets of simulated PET/CT SPN images. The first subset is currently composed of 1000 cases containing pairs of the transaxial CT and the corresponding PET slice with various types of simulated SPNs, presented as individual records. The second subset contains pairs of the transaxial CT and the corresponding PET slice of simulated SPNs, presenting cases of graded difficulty in diagnosis. The users of the database will have the ability to set queries in order to retrieve cases with certain characteristics, as well as characterized image sets. All images are freely available and may be downloaded from the website. SPNsim provides a useful reference data set for training and evaluation of computer aided detection (CAD) and diagnosis (CADx) systems focusing on SPN.

A methodology for the estimation of the optimal iteration in MLEM-based image reconstruction in PET

We have proposed a method for the estimation of the optimal iteration for the Maximum Likelihood Expectation Maximization (MLEM) algorithm used in Positron Emission Tomography (PET) image reconstruction. For the calculation of the transition matrix and the generation of the projection data a PET scanner was simulated using Monte-Carlo techniques. Our preliminary results show that the proposed methodology can estimate the optimal iteration for the MLEM algorithm.