Segmentation method for breast tumor diagnosis based on Artificial Neural Network algorithm applied to dynamic 18F-FDG PET images

Standard 3D dynamic positron emission tomographic (PET) imaging consists of independent image reconstructions of individual frames followed by application of appropriate kinetic model to the time activity curves at the voxel or region-of-interest (ROI). The emerging field of 4D PET reconstruction, by contrast, seeks to move beyond this scheme and incorporate information from multiple frames within the image reconstruction task. Here we propose a novel reconstruction framework aiming to enhance quantitative accuracy of parametric images via introduction of priors based on voxel kinetics, as generated via clustering of preliminary reconstructed dynamic images to define clustered neighborhoods of voxels with similar kinetics. This is then followed by straightforward maximum a posteriori (MAP) 3D PET reconstruction as applied to individual frames; and as such the method is labeled ‘3.5D’ image reconstruction. The use of cluster-based priors has the advantage of further enhancing quantitative performance in dynamic PET imaging, because: (a) there are typically more voxels in clusters than in conventional local neighborhoods, and (b) neighboring voxels with distinct kinetics are less likely to be clustered together. Using realistic simulated 11C-raclopride dynamic PET data, the quantitative performance of the proposed method was investigated. Parametric distribution-volume (DV) and DV ratio (DVR) images were estimated from dynamic image reconstructions using (a) maximum-likelihood expectation maximization (MLEM), and MAP reconstructions using (b) the quadratic prior (QP-MAP), (c) the Green prior (GP-MAP) and (d, e) two proposed cluster-based priors (CP-U-MAP and CP-W-MAP), followed by graphical modeling, and were qualitatively and quantitatively compared for 11 ROIs. Overall, the proposed dynamic PET reconstruction methodology resulted in substantial visual as well as quantitative accuracy improvements (in terms of noise versus bias performance) for parametric DV and DVR images. The method was also tested on a 90 min 11C-raclopride patient study performed on the high-resolution research tomography. The proposed method was shown to outperform the conventional method in visual as well as quantitative accuracy improvements (in terms of noise versus regional DVR value performance).

Dynamic positron emission tomography (PET) enables the quantification of physiological parameters of radiotracers employed in the investigation of neuropsychiatric disorders. We previously introduced a factor analysis-based algorithm, Cluster-Initialized Factor Analysis (CIFA), designed to overcome the problem of specifying reference regions. CIFA is capable of automatically extracting distinct radiotracer binding distributions across many modalities based on the differences in tracer dynamics, and thus can distinguish regions of specific- and non-specific binding without requiring prior segmentation. Our goal is to quantitatively validate the ability of CIFA to resolve different dynamic biological processes by comparing the output of the algorithm to an independent benchmark. As an intermediate goal, we aim to create a physical phantom capable of modeling unique aspects of dynamic imaging and to use this phantom as the benchmark in evaluatingCIFA. CIFA was used to reconstruct 18F-flortaucipir dynamic brain PET datasets acquired at Lawrence Berkeley National Lab. The resulting factor curves served as the foundation for creating dynamic input time-activity curve (TAC) combinations in a physical brain phantom specifically constructed for this purpose. The phantom represented three components: two overlapping tissue types and free radiotracer, constructed with a combination of small hydraulic elements. The physical components were scanned separately to generate a library of images, allowing us to reproduce scans of any duration with prescribed dynamics and realistic partial volume effects. The phantom was designed to produce noisy instances with compartment mixing of dynamic scans with desired activity TACs for free, non-specifically bound, and specifically bound radiotracers. Ten distinct dynamic simulations with varying levels of TAC similarity were estimated withCIFA. We directly evaluated CIFA's performance in analyzing each of the 10 dynamic datasets by computing the Pearson correlation coefficient between the estimated outputs and the ground truth tissue TACs and corresponding tissue distributions. For seven out of 10 modeled dynamics, which captured the full spectrum of realistically expected tissue TAC shapes, the curve correlation of the specific binding tissue was above 95%. This work formulated an innovative process by combining a physical phantom design with PET images for evaluating the application of CIFA in the extraction of dynamic TACs from dynamic PET image data. In most cases the CIFA algorithm accurately reproduced the dynamics of the phantom simulateddata.

Positron emission tomography (PET) images are corrupted by noise. This is especially true in dynamic PET imaging where short frames are required to capture the peak of activity concentration after the radiotracer injection. High noise results in a possible bias in quantification, as the compartmental models used to estimate the kinetic parameters are sensitive to noise. This paper describes a new post-reconstruction filter to increase the signal-to-noise ratio in dynamic PET imaging. It consists in a spatio-temporal robust diffusion of the 4D image based on the time activity curve (TAC) in each voxel. It reduces the noise in homogeneous areas while preserving the distinct kinetics in regions of interest corresponding to different underlying physiological processes. Neither anatomical priors nor the kinetic model are required. We propose an automatic selection of the scale parameter involved in the diffusion process based on a robust statistical analysis of the distances between TACs. The method is evaluated using Monte Carlo simulations of brain activity distributions. We demonstrate the usefulness of the method and its superior performance over two other post-reconstruction spatial and temporal filters. Our simulations suggest that the proposed method can be used to significantly increase the signal-to-noise ratio in dynamic PET imaging.

Small-animal positron emission tomography as a tool for neuropharmacology

PurposeThe glutamine analogue (2S, 4R)-4-[18F]fluoroglutamine ([18F]FGln) was investigated to further characterize its pharmacokinetics and acquire in vivo positron emission tomography (PET) images of separate orthotopic and subcutaneous glioma xenografts in mice.Procedures[18F]FGln was synthesized at a high radiochemical purity as analyzed by high-performance liquid chromatography. An orthotopic model was created by injecting luciferase-expressing patient-derived BT3 glioma cells into the right hemisphere of BALB/cOlaHsd-Foxn1nu mouse brains (tumor growth monitored via in vivo bioluminescence), the subcutaneous model by injecting rat BT4C glioma cells into the flank and neck regions of Foxn1nu/nu mice. Dynamic PET images were acquired after injecting 10–12 MBq of the tracer into mouse tail veins. Animals were sacrificed 63 min after tracer injection, and ex vivo biodistributions were measured. Tumors and whole brains (with tumors) were cryosectioned, autoradiographed, and stained with hematoxylin-eosin. All images were analyzed with CARIMAS software. Blood sampling of 6 Foxn1nu/nu and 6 C57BL/6J mice was performed after 9–14 MBq of tracer was injected at time points between 5 and 60 min then assayed for erythrocyte uptake, plasma protein binding, and plasma parent-fraction of radioactivity to correct PET image-derived whole-blood radioactivity and apply the data to multiple pharmacokinetic models.ResultsOrthotopic human glioma xenografts displayed PET image tumor-to-healthy brain region ratio of 3.6 and 4.8 while subcutaneously xenografted BT4C gliomas displayed (n = 12) a tumor-to-muscle (flank) ratio of 1.9 ± 0.7 (range 1.3–3.4). Using PET image-derived blood radioactivity corrected by population-based stability analyses, tumor uptake pharmacokinetics fit Logan and Yokoi modeling for reversible uptake.ConclusionsThe results reinforce that [18F]FGln has preferential uptake in glioma tissue versus that of corresponding healthy tissue and fits well with reversible uptake models.

Dynamic total-body positron emission tomography (PET) imaging has recently become possible due to the introduction of new-generation imaging systems with increased axial field of view. This enables evaluation of kinetic modeling approaches using compartment models for the entire body simultaneously. Systematic evaluation of the conventionally used approaches is needed to find the best candidates among different compartment models for total body modeling of 15O-water perfusion PET images. Six variations of the 1-tissue-compartment model (1TCM) are evaluated. The models are systematically fitted to mean time activity curves of 20 anatomic structures of interest extracted from a dynamic 15O-water PET rest perfusion images of 58 human subjects, using an image-derived input function from the aorta. The impact of additional parameters for time delay correction (TDC), arterial volume fraction (AVF), and perfusable tissue fraction (PTF) is assessed. The models are evaluated with mean squared error, mean relative error (MRE), and Akaike information criterion (AIC) and by performing Wilcoxon signed-rank tests to see whether the found differences are statistically significant. For 11 out of the 20 anatomic structures, the best model in terms of median AIC was the 1TCM with AVF and TDC but without PTF. The use of TDC resulted in significantly lower AICs for all anatomic structures while the AVF and PTF parameters significantly improved the model performance only occasionally. Between the different organs, the models of pancreas, spleen, myocardium, the example rib, and brain performed very well in terms of MRE. In comparison, the investigated models worked poorly for liver and lung lobes. This paper serves as an initial attempt to evaluate potential approaches to derive an appropriate model for kinetic modeling in total body 15O-water PET perfusion imaging. According to our research, the suitability of 1TCM using a single input function from the aorta in 15O-water PET images depends on the organ of interest. Our results suggest that the use of 1TCM is justified for most anatomic structures and can be used to estimate the cerebral, myocardial, renal, and splenic blood flow. However, we recommend different models for quantification of hepatic and pulmonary blood flow.

In order to determine if dynamic positron emission tomography (PET) and N-13 ammonia can be used to quantitate regional RBF (rRBF) noninvasively, six anesthetized dogs were examined with PET imaging after an iv bolus administration of 5 mCi of N-13 ammonia. Renal time activity curves and the arterial input function were derived from regions of interest drawn over the renal cortex and abdominal aorta, respectively. For calculation of rRBF, less than 120 s of the initial data were used to minimize contamination by plasma metabolites of N-13 radioactivity. rRBF was quantitated with a two-compartment model, and the results were compared with simultaneously acquired microsphere blood flow measurement. Fourteen experiments were performed in six dogs, and four regions of interest on renal cortex were selected on each PET image. RBF derived from dynamic PET imaging with N-13 ammonia was linearly related to microsphere (MS) values (rRBF = 1.06 x MS - 0.17; r = 0.91). Mean rRBF in the canine experiments was 4.0 mL/min/g. The results indicate that dynamic N-13 ammonia renal PET can provide noninvasively quantitative rRBF.

Quantitative evaluation of the improvement in the pharmacokinetics of a nucleic acid drug delivery system by dynamic PET imaging with 18F-incorporated oligodeoxynucleotides

Optical imaging has been the primary imaging modality for nearly all of the renal clearable nanoparticles since 2007. Due to the tissue depth penetration limitation, providing accurate organ kinetics non-invasively has long been a huge challenge. Although a more quantitative imaging technique has been developed by labeling nanoparticles with single-photon emission computed tomography (SPECT) isotopes, the low temporal resolution of SPECT still limits its potential for visualizing the rapid dynamic process of renal clearable nanoparticles in vivo. The dynamic positron emission tomography (PET) imaging of renal clearable gold (Au) nanoparticles by labeling them with copper-64 ((64) Cu) to form (64) Cu-NOTA-Au-GSH is reported. Systematic nanoparticle synthesis and characterizations are performed to demonstrate the efficient renal clearance of as-prepared nanoparticles. A rapid renal clearance of (64) Cu-NOTA-Au-GSH is observed (>75%ID at 24 h post-injection) with its elimination half-life calculated to be less than 6 min, over 130 times shorter than previously reported similar nanoparticles. Dynamic PET imaging not only addresses the current challenges in accurately and non-invasively acquiring the organ kinetics, but also potentially provides a highly useful tool for studying renal clearance mechanism of other ultra-small nanoparticles, as well as the diagnosis of kidney diseases in the near future.

Comparison of time curves from dynamic 18F-fluciclovine positron emission tomography and dynamic contrast-enhanced magnetic resonance imaging for primary prostate carcinomas

Respiratory motion blurs positron emission tomography (PET) images in thorax and abdominal regions, and can cause attenuation-related errors in quantitative parameters. In addition, this motion can cause localization error and disappearance of tumor(s) near the diaphragm. Respiratory gated PET imaging, or 4D PET imaging, is an attractive solution. For 4D PET, a surrogate measurement of the patient's breathing is made during the scan, and based on this measurement the list-mode PET data are sorted and reconstructed into multiple images. In several studies, it has been reported that 4D PET imaging recovers tumor volume more accurately than 3D PET imaging, and yields more accurate standard uptake values. In spite of its promise, 4D PET images suffer from statistical noise because only a fraction of the acquired data can be used in each image. Moreover, they also suffer from inconsistent attenuation correction due to mismatches between PET and CT images. In this paper, we propose a novel framework for phased attenuation correction and respiratory motion compensation for a 3D PET image by using a 3D CT image and multiple respiratory-phase 3D MR images. For phased attenuation correction, we generate and utilize 4D CT image obtained via non-rigid registrations among a 3D CT image and multiple respiratory-phase 3D MR images. A 3D PET image at a selected respiratory phase is then obtained by adding attenuation corrected multiple 3D PET images through non-rigid transformation. Experimental result for a clinical dataset shows that the proposed algorithm can provide much clearer organ boundary with improved quality than the conventional method in a 3D PET image.

Positron emission tomography (PET) is a non-invasive molecular imaging method for quantitative observation of physiological and biochemical changes in living organisms. The quality of the reconstructed PET image is limited by many different physical degradation factors. Various denoising methods including Gaussian filtering (GF) and non-local mean (NLM) filtering have been proposed to improve the image quality. However, image denoising usually blurs edges, of which high frequency components are filtered as noises. On the other hand, it is well-known that edges in a PET image are important to detection and recognition of a lesion. Denoising while preserving the edges of PET images remains an important yet challenging problem in PET image processing. In this paper, we propose a novel denoising method with good edge-preserving performance based on spectral graph wavelet transform (SGWT) for dynamic PET images denoising. We firstly generate a composite image from the entire time series, then perform SGWT on the PET images, and finally reconstruct the low graph frequency content to get the denoised dynamic PET images. Experimental results on simulation and in vivo data show that the proposed approach significantly outperforms the GF, NLM and graph filtering methods. Compared with deep learning-based method, the proposed method has the similar denoising performance, but it does not need lots of training data and has low computational complexity.

Positron emission tomography (PET) imaging of activated T-cells with N-(4-[18F]fluorobenzoyl)-interleukin-2 ([18F]FB-IL-2) may be a promising tool for patient management to aid in the assessment of clinical responses to immune therapeutics. Unfortunately, existing radiosynthetic methods are very low yielding due to complex and time-consuming chemical processes. Herein, we report an improved method for the synthesis of [18F]FB-IL-2, which reduces synthesis time and improves radiochemical yield. With this optimized approach, [18F]FB-IL-2 was prepared with a non-decay-corrected radiochemical yield of 3.8 ± 0.7% from [18F]fluoride, 3.8 times higher than previously reported methods. In vitro experiments showed that the radiotracer was stable with good radiochemical purity (>95%), confirmed its identity and showed preferential binding to activated mouse peripheral blood mononuclear cells. Dynamic PET imaging and ex vivo biodistribution studies in naïve Balb/c mice showed organ distribution and kinetics comparable to earlier published data on [18F]FB-IL-2. Significant improvements in the radiochemical manufacture of [18F]FB-IL-2 facilitates access to this promising PET imaging radiopharmaceutical, which may, in turn, provide useful insights into different tumour phenotypes and a greater understanding of the cellular nature and differential immune microenvironments that are critical to understand and develop new treatments for cancers.

Dynamic positron emission tomography (PET) imaging provides important quantitative information of physiological and biochemical processes in humans and animals. However, due to short-time acquisitions to obtain a time sequence of images for parametric imaging, the signal-to-noise ratio of measurement data in each time frame is often very low, which leads the dynamic PET image reconstruction to be a challenging task. And some noticeable errors are inevitable transferred to the voxel-wise kinetic parameter imaging from the associative noisy TAC measurements. To tackle this problem, maximum a posteriori (MAP) statistical reconstruction methods are widely used by incorporating some prior information. Conventional priors focus on local neighborhoods in individual image frames and subsequently penalize inter-voxel intensity differences through different penalty functions such as the quadratic membrane smoothing prior and non-quadratic edge-preserving prior, failing to explore the temporal information of dynamic PET data. In this paper, we design a spatial-temporal edge-preserving (STEP) prior model under the framework of bilateral filter by considering both the spatial local neighborhoods and the temporal kinetic information. Experimental results via a computer simulation study demonstrate that the present dynamic PET reconstruction method with the STEP prior can achieve noticeable gains than the conventional Huber prior in term of signal-to-noise and bias-variance evaluations for the parametric images.

The aim of this study is to develop and evaluate a novel direct reconstruction method to improve the signal-to-noise ratio (SNR) of parametric images in dynamic positron-emission tomography (PET), especially for applications in myocardial perfusion studies. Simulation studies were used to test the performance in SNR and computational efficiency for different methods. The NCAT phantom was used to generate simulated dynamic data. Noise realization was performed in the sinogram domain and repeated for 30 times with four different noise levels by varying the injection dose (ID) from standard ID to 1/8 of it. The parametric images were calculated by (1) three direct methods that compute the kinetic parameters from the sinogram and (2) an indirect method, which computes the kinetic parameter with pixel-by-pixel curve fitting in image space using weighted least-squares. The first direct reconstruction maximizes the likelihood function using trust-region-reflective (TRR) algorithm. The second approach uses tabulated parameter sets to generate precomputed time-activity curves for maximizing the likelihood functions. The third approach, as a newly proposed method, assumes separable complete data to derive the M-step for maximizing the likelihood. The proposed method with the separable complete data performs similarly to the other two direct reconstruction methods in terms of the SNR, providing a 5%-10% improvement as compared to the indirect parametric reconstruction under the standard ID. The improvement of SNR becomes more obvious as the noise level increases, reaching more than 30% improvement under 1/8 ID. Advantage of the proposed method lies in the computation efficiency by shortening the time requirement to 25% of the indirect approach and 3%-6% of other direct reconstruction methods. With results provided from this simulation study, direct reconstruction of myocardial blood flow shows a high potential for improving the parametric image quality for clinical use.