Non-Local Means Denoising of Dynamic PET Images

Objective

Dynamic positron emission tomography (PET), which reveals information about both the spatial distribution and temporal kinetics of a radiotracer, enables quantitative interpretation of PET data. Model-based interpretation of dynamic PET images by means of parametric fitting, however, is often a challenging task due to high levels of noise, thus necessitating a denoising step. The objective of this paper is to develop and characterize a denoising framework for dynamic PET based on non-local means (NLM).

Theory

NLM denoising computes weighted averages of voxel intensities assigning larger weights to voxels that are similar to a given voxel in terms of their local neighborhoods or patches. We introduce three key modifications to tailor the original NLM framework to dynamic PET. Firstly, we derive similarities from less noisy later time points in a typical PET acquisition to denoise the entire time series. Secondly, we use spatiotemporal patches for robust similarity computation. Finally, we use a spatially varying smoothing parameter based on a local variance approximation over each spatiotemporal patch.

Methods

To assess the performance of our denoising technique, we performed a realistic simulation on a dynamic digital phantom based on the Digimouse atlas. For experimental validation, we denoised PET images from a mouse study and a hepatocellular carcinoma patient study. We compared the performance of NLM denoising with four other denoising approaches – Gaussian filtering, PCA, HYPR, and conventional NLM based on spatial patches.

Results

The simulation study revealed significant improvement in bias-variance performance achieved using our NLM technique relative to all the other methods. The experimental data analysis revealed that our technique leads to clear improvement in contrast-to-noise ratio in Patlak parametric images generated from denoised preclinical and clinical dynamic images, indicating its ability to preserve image contrast and high intensity details while lowering the background noise variance.     

Short-duration dynamic FDG PET imaging: Optimization and clinical application

PurposeWe aimed to investigate whether short dynamic PET imaging started at injection, complemented with routine clinical acquisition at 60-min post-injection (static), can achieve reliable kinetic analysis.MethodsDynamic and static 18F-2-fluoro-2-deoxy-D-glucose (FDG) PET data were generated using realistic simulations to assess uncertainties due to statistical noise as well as bias. Following image reconstructions, kinetic parameters obtained from a 2-tissue-compartmental model (2TCM) were estimated, making use of the static image, and the time duration of dynamic PET data were incrementally shortened. We also investigated, in the first 2-min, different frame sampling rates, towards optimized dynamic PET imaging. Kinetic parameters from shortened dynamic datasets were additionally estimated for 9 patients (15 scans) with liver metastases of colorectal cancer, and were compared with those derived from full dynamic imaging using correlation and Passing–Bablok regression analyses.ResultsThe results showed that by reduction of dynamic scan times from 60-min to as short as 5-min, while using static data at 60-min post-injection, bias and variability stayed comparable in estimated kinetic parameters. Early frame samplings of 5, 24 and 30 s yielded highest biases compared to other schemes. An early frame sampling of 10 s generally kept both bias and variability to a minimum. In clinical studies, strong correlation (r ≥ 0.97, P < 0.0001) existed between all kinetic parameters in full vs. shortened scan protocols.ConclusionsShortened 5-min dynamic scan, sampled as 12 × 10 + 6 × 30 s, followed by 3-min static image at 60-min post-injection, enables accurate and robust estimation of 2TCM parameters, while enabling generation of SUV estimates.     

Forging a path to mesoscopic imaging success with ultra-high field functional magnetic resonance imaging

Functional magnetic resonance imaging (fMRI) studies with ultra-high field (UHF, 7+ Tesla) technology enable the acquisition of high-resolution images. In this work, we discuss recent achievements in UHF fMRI at the mesoscopic scale, on the order of cortical columns and layers, and examine approaches to addressing common challenges. As researchers push to smaller and smaller voxel sizes, acquisition and analysis decisions have greater potential to degrade spatial accuracy, and UHF fMRI data must be carefully interpreted. We consider the impact of acquisition decisions on the spatial specificity of the MR signal with a representative dataset with 0.8 mm isotropic resolution. We illustrate the trade-offs in contrast with noise ratio and spatial specificity of different acquisition techniques and show that acquisition blurring can increase the effective voxel size by as much as 50% in some dimensions. We further describe how different sources of degradations to spatial resolution in functional data may be characterized. Finally, we emphasize that progress in UHF fMRI depends not only on scientific discovery and technical advancement, but also on informal discussions and documentation of challenges researchers face and overcome in pursuit of their goals.This article is part of the theme issue ‘Key relationships between non-invasive functional neuroimaging and the underlying neuronal activity’.     

A rigid body model of the forearm

In this article the forearm, with its complex, continuous motion of masses during pronation/supination, was approximated by a rigid body model consisting of a radial segment rotating around an ulnar segment. The method used to obtain the model parameters is based on three-dimensional voxel data that include velocity information. We propose a criterion that allows the voxels to be attributed to either of the two segments. It is based on the notion that the rotational kinetic energy determined from the voxel data equals the kinetic energy of the rigid body model. To obtain a three-dimensional smoothing we further propose a parameterization of the shape of both segments. These shapes can then be used to determine the dynamic integrals of the segments, i.e. mass, center of mass, and inertia. Using this approach we determined all model parameters for a human forearm from three series of MRI scans in a supinated, a pronated, and an intermediate position. In the appendix, a procedure is described that allows the dynamic quantities to be scaled homogeneously without recalculation of the integrals. Thus, this article provides all essential parameters required for three-dimensional dynamic simulations of general movements of the forearm.     

Applications cliniques de l’imagerie hybride TEP-IRM

This paper reports the results of a preliminary study evaluating the feasibility and performance of a first whole body hybrid PET/MR scanner allowing sequential acquisition of co-registered MR and PET images. Sixty-two patients underwent whole body PET/MR imaging immediately after a clinical PET/CT. The hybrid device consists of a 3T MR and a time-of-flight PET scanner sharing a single bed allowing sequential acquisition of co-registered MR and PET images. Imaging protocols included a whole body MR used for attenuation correction of PET followed by high-resolution diagnostic MR. Image analysis included visual identification of radiotracer uptake in tumors and measurement of standardized uptake values (SUV) in tumoral lesions and in normal organs. PET images acquired in the PET/MR with a delay of 85 ± 22 minutes (range 49–120 minutes) showed perfect correlation and identical diagnostic quality compared to PET/CT. In 42 patients (68%), additional high-resolution MR sequences were acquired for clinical diagnosis showing excellent quality without any visually detectable artifacts. SUV measurements of tumor lesions obtained after correction with MR attenuation maps showed an excellent correlation with PET/CT (R² = 0.89 and R² = 0.95 for mean and maximum tissue uptake respectively). Due to the delay between the two studies, changes in tracer uptake biodistribution of normal tissue were observed. Our preliminary data show that whole body PET/MR is clinically applicable in oncologic patients yielding a comparable diagnostic performance as PET/CT with respect to lesion detection and localization.     

TEP/IRM hybride en neuro-imagerie

The hybrid Positron Emission Tomography/Magnetic Resonance Imaging (PET/MRI) is a newly available imaging modality combining the molecular and metabolic PET information with the morphological and functional data provided by MRI. Integrated PET/MRI tomographs were conceived in analogy to the current PET/Computed Tomography (PET/CT) technology, with specific properties linked to the intrinsic differences of MRI and CT imaging. In the field of neuro-imaging, in particular, MRI provides a larger panel of information, as compared with CT, and is already systematically fused and used as a support for PET images for diagnostic and research purposes. We summarize here our current experience with the first integrated PET/MRI tomograph installed in Switzerland, concerning specifically three clinical applications: brain tumors characterization, the diagnosis of neurodegenerative dementias and the presurgical evaluation of pharmaco-resistant epilepsy. With this sequential tomograph, we could combine the full range of diagnostic MR sequences (including diffusion tensor imaging, tractography, spectroscopy, functional MR) with PET imaging of brain glucose metabolism (by ¹⁸F-Fluorodeoxyglucose–FDG) and of amino acid transport (by ¹⁸F-Fluoroethyltyrosine–FET). We also summarize the main results obtained in neuro-imaging by the different groups working with these new hybrid tomographs. These data show that PET/MRI, acquired in a single imaging session, may represent the modality of choice for neuro-imaging.     

Joint SFMN/ANOCEF focus on 18F-FDOPA PET imaging in glioma: Current applications and perspectives

18F-FDOPA PET has demonstrated its additional value during the clinical course of glioma, at initial diagnosis, for treatment planning or follow-up. The aim of the current review was to summarize current applications of 18F-FDOPA PET in gliomas and constitute, as a perspective, a first step in harmonizing clinical practices in French centers. In France, the indication for 18F-FDOPA PET is restricted to the assessment of primary brain tumor recurrence. According to the literature, this indication could be expanded to primary diagnosis and, to a lesser extent, treatment monitoring. There is a real need to harmonize standard procedures among French centers. The objective is to increase the availability of data for this rare entity of glioma and to develop multi-parametric PET analyses (static, dynamic and textural), also known as radiomics, by using artificial intelligence algorithms. For this purpose, kinetics analysis with dynamic PET acquisition should be implemented in routine practice because it has demonstrated its additional value for initial diagnosis in gliomas. Therefore, this review proposes a workflow based on acquisition and reconstruction parameters that can be implemented in each center to increase the amount of standardized 18F-FDOPA PET data in neuro-oncology imaging in France. This would help in creating a national database and developing national multi-center studies that can respond to the challenge of using multi-parametric PET in glioma.     

Benefits and risks of transforming data from dynamic positron emission tomography, with an application to hepatic encephalopathy

Transforming data sets to bring out expected model features can be valuable within limits and misleading outside them. Here we establish such limits for the widely used Gjedde-Patlak representation of dynamic PET data, with an application to hepatic encephalopathy.     

Assessing skeletal muscle glucose metabolism with positron emission tomography

Insulin has a marked effect to stimulate the transport and metabolism of glucose in skeletal muscle in healthy individuals, whereas an impaired response, termed insulin resistance, is a major risk factor for diabetes mellitus and other metabolic diseases. Studies of the molecular physiology of insulin action in skeletal muscle indicate that a principal loci of control resides within the proximal steps of glucose transport and phosphorylation. Deoxyglucose, the metabolism of which is limited to these proximal steps, is widely used for in vitro studies of insulin action on glucose transport. The technologies of PET imaging provide a unique opportunity to carry out similar studies in vivo in human skeletal muscle. In this instance, a short-lived positron labeled tracer, [18F] FDG, can be given at sufficiently high specific activity to image not only glucose uptake, but by dynamic PET imaging, by monitoring the time course of [18F] FDG tissue activity, data can be generated to examine the kinetics of glucose transport and phosphorylation. The experimental procedures of this approach, including an overview of the mathematical modeling, are described in this review, along with some of the key findings of the initial applications of PET for the study of glucose metabolism in human skeletal muscle.     

Automated Movement Correction for Dynamic PET/CT Images: Evaluation with Phantom and Patient Data

Head movement during a dynamic brain PET/CT imaging results in mismatch between CT and dynamic PET images. It can cause artifacts in CT-based attenuation corrected PET images, thus affecting both the qualitative and quantitative aspects of the dynamic PET images and the derived parametric images. In this study, we developed an automated retrospective image-based movement correction (MC) procedure. The MC method first registered the CT image to each dynamic PET frames, then re-reconstructed the PET frames with CT-based attenuation correction, and finally re-aligned all the PET frames to the same position. We evaluated the MC method''s performance on the Hoffman phantom and dynamic FDDNP and FDG PET/CT images of patients with neurodegenerative disease or with poor compliance. Dynamic FDDNP PET/CT images (65 min) were obtained from 12 patients and dynamic FDG PET/CT images (60 min) were obtained from 6 patients. Logan analysis with cerebellum as the reference region was used to generate regional distribution volume ratio (DVR) for FDDNP scan before and after MC. For FDG studies, the image derived input function was used to generate parametric image of FDG uptake constant (K_i) before and after MC. Phantom study showed high accuracy of registration between PET and CT and improved PET images after MC. In patient study, head movement was observed in all subjects, especially in late PET frames with an average displacement of 6.92 mm. The z-direction translation (average maximum = 5.32 mm) and x-axis rotation (average maximum = 5.19 degrees) occurred most frequently. Image artifacts were significantly diminished after MC. There were significant differences (P<0.05) in the FDDNP DVR and FDG Ki values in the parietal and temporal regions after MC. In conclusion, MC applied to dynamic brain FDDNP and FDG PET/CT scans could improve the qualitative and quantitative aspects of images of both tracers.     

Bayesian spatiotemporal modeling on complex-valued fMRI signals via kernel convolutions

We propose a model-based approach that combines Bayesian variable selection tools, a novel spatial kernel convolution structure, and autoregressive processes for detecting a subject's brain activation at the voxel level in complex-valued functional magnetic resonance imaging (CV-fMRI) data. A computationally efficient Markov chain Monte Carlo algorithm for posterior inference is developed by taking advantage of the dimension reduction of the kernel-based structure. The proposed spatiotemporal model leads to more accurate posterior probability activation maps and less false positives than alternative spatial approaches based on Gaussian process models, and other complex-valued models that do not incorporate spatial and/or temporal structure. This is illustrated in the analysis of simulated data and human task-related CV-fMRI data. In addition, we show that complex-valued approaches dominate magnitude-only approaches and that the kernel structure in our proposed model considerably improves sensitivity rates when detecting activation at the voxel level.     

The Problem of Thresholding in Small-World Network Analysis

Graph theory deterministically models networks as sets of vertices, which are linked by connections. Such mathematical representation of networks, called graphs are increasingly used in neuroscience to model functional brain networks. It was shown that many forms of structural and functional brain networks have small-world characteristics, thus, constitute networks of dense local and highly effective distal information processing. Motivated by a previous small-world connectivity analysis of resting EEG-data we explored implications of a commonly used analysis approach. This common course of analysis is to compare small-world characteristics between two groups using classical inferential statistics. This however, becomes problematic when using measures of inter-subject correlations, as it is the case in commonly used brain imaging methods such as structural and diffusion tensor imaging with the exception of fibre tracking. Since for each voxel, or region there is only one data point, a measure of connectivity can only be computed for a group. To empirically determine an adequate small-world network threshold and to generate the necessary distribution of measures for classical inferential statistics, samples are generated by thresholding the networks on the group level over a range of thresholds. We believe that there are mainly two problems with this approach. First, the number of thresholded networks is arbitrary. Second, the obtained thresholded networks are not independent samples. Both issues become problematic when using commonly applied parametric statistical tests. Here, we demonstrate potential consequences of the number of thresholds and non-independency of samples in two examples (using artificial data and EEG data). Consequently alternative approaches are presented, which overcome these methodological issues.     

Modeling Inter‐Subject Variability in fMRI Activation Location: A Bayesian Hierarchical Spatial Model

Summary The aim of this article is to develop a spatial model for multi‐subject fMRI data. There has been extensive work on univariate modeling of each voxel for single and multi‐subject data, some work on spatial modeling of single‐subject data, and some recent work on spatial modeling of multi‐subject data. However, there has been no work on spatial models that explicitly account for inter‐subject variability in activation locations. In this article, we use the idea of activation centers and model the inter‐subject variability in activation locations directly. Our model is specified in a Bayesian hierarchical framework which allows us to draw inferences at all levels: the population level, the individual level, and the voxel level. We use Gaussian mixtures for the probability that an individual has a particular activation. This helps answer an important question that is not addressed by any of the previous methods: What proportion of subjects had a significant activity in a given region. Our approach incorporates the unknown number of mixture components into the model as a parameter whose posterior distribution is estimated by reversible jump Markov chain Monte Carlo. We demonstrate our method with a fMRI study of resolving proactive interference and show dramatically better precision of localization with our method relative to the standard mass‐univariate method. Although we are motivated by fMRI data, this model could easily be modified to handle other types of imaging data.     

对比剂动力学时间分辨成像及弥散加权成像技术活体动态监测兔VX2肌肉肿瘤生物学生长与血管生成

目的:应用对比剂动力学时间分辨成像(Time Resolved Imaging of Contrast Kinetics,TRICKS)技术增强磁共振血管成像(MRangiography,MRA)及弥散加权成像(Diffusion Weighted Imaging,DWI)技术活体动态监测兔VX2肌肉肿瘤生物学生长与血管生成,探讨肿瘤血管生成与肿瘤生长的关系。方法:30只新西兰白兔,每只均在右后腿肌肉内接种VX2肿瘤细胞1×107建立肿瘤模型。分别在肿瘤接种后第4、7、10、13、16天(每个时间点6只)分别进行T1WI、T2WI、DWI、TRICKS动态增强MRA及T1WI增强延迟扫描,活体监测兔VX2肌肉肿瘤血管生成,肿瘤标本HE及CD31免疫组化染色进行验证。两位医师双盲法分别测量不同生长点肿瘤的长、短径及体积,并与大体病理标本比较;测定TRICKS增强动态MRA所能显示肿瘤血管的最小直径及形态变化;观察ADC值变化与肿瘤生长的关系。结果(:1)ADC值随着肿瘤体积的长大而逐渐增大。(2)MRI活体测定肿瘤大小与病理大体标本所测算肿瘤体积的差异无显著性。(3)TRICKS增强MRA动态显示肿瘤血管的最小...     

Assessment of Noninferiority (and Equivalence) for Simple Crossover Trials Using Bayesian Approach

To assess the noninferiority or equivalence of a general drug to a standard one, researchers generally use the odds ratio of patient response rates to evaluate the relative treatment efficacy. In this paper, we use a logistic random effects model to test noninferiority and equivalence based on the odds ratio of patient response rates for crossover trials with binary data. We use Bayesian sampling algorithm, data augmentation, and scaled mixture of normal representation to implement the approach and improve efficiency. The performance of the proposed approach is assessed via simulation and real data examples.     

Bayesian hierarchically weighted finite mixture models for samples of distributions

Finite mixtures of Gaussian distributions are known to provide an accurate approximation to any unknown density. Motivated by DNA repair studies in which data are collected for samples of cells from different individuals, we propose a class of hierarchically weighted finite mixture models. The modeling framework incorporates a collection of k Gaussian basis distributions, with the individual-specific response densities expressed as mixtures of these bases. To allow heterogeneity among individuals and predictor effects, we model the mixture weights, while treating the basis distributions as unknown but common to all distributions. This results in a flexible hierarchical model for samples of distributions. We consider analysis of variance-type structures and a parsimonious latent factor representation, which leads to simplified inferences on non-Gaussian covariance structures. Methods for posterior computation are developed, and the model is used to select genetic predictors of baseline DNA damage, susceptibility to induced damage, and rate of repair.     

Creating Dynamic Images of Short-lived Dopamine Fluctuations with lp-ntPET: Dopamine Movies of Cigarette Smoking

We describe experimental and statistical steps for creating dopamine movies of the brain from dynamic PET data. The movies represent minute-to-minute fluctuations of dopamine induced by smoking a cigarette. The smoker is imaged during a natural smoking experience while other possible confounding effects (such as head motion, expectation, novelty, or aversion to smoking repeatedly) are minimized.We present the details of our unique analysis. Conventional methods for PET analysis estimate time-invariant kinetic model parameters which cannot capture short-term fluctuations in neurotransmitter release. Our analysis - yielding a dopamine movie - is based on our work with kinetic models and other decomposition techniques that allow for time-varying parameters ^1-7. This aspect of the analysis - temporal-variation - is key to our work. Because our model is also linear in parameters, it is practical, computationally, to apply at the voxel level. The analysis technique is comprised of five main steps: pre-processing, modeling, statistical comparison, masking and visualization. Preprocessing is applied to the PET data with a unique ''HYPR'' spatial filter ⁸ that reduces spatial noise but preserves critical temporal information. Modeling identifies the time-varying function that best describes the dopamine effect on ¹¹C-raclopride uptake. The statistical step compares the fit of our (lp-ntPET) model ⁷ to a conventional model ⁹. Masking restricts treatment to those voxels best described by the new model. Visualization maps the dopamine function at each voxel to a color scale and produces a dopamine movie. Interim results and sample dopamine movies of cigarette smoking are presented.