Machine Learning for the Design and the Simulation of Radiofrequency Magnetic Resonance Coils: Literature Review, Challenges, and Perspectives.

Radiofrequency (RF) coils for magnetic resonance imaging (MRI) applications serve to generate RF fields to excite the nuclei in the sample (transmit coil) and to pick up the RF signals emitted by the nuclei (receive coil). For the purpose of optimizing the image quality, the performance of RF coils has to be maximized. In particular, the transmit coil has to provide a homogeneous RF magnetic field, while the receive coil has to provide the highest signal-to-noise ratio (SNR). Thus, particular attention must be paid to the coil simulation and design phases, which can be performed with different computer simulation techniques. Being largely used in many sectors of engineering and sciences, machine learning (ML) is a promising method among the different emerging strategies for coil simulation and design. Starting from the applications of ML algorithms in MRI and a short description of the RF coil's performance parameters, this narrative review describes the applications of such techniques for the simulation and design of RF coils for MRI, by including deep learning (DL) and ML-based algorithms for solving electromagnetic problems.

Deep Learning Staging of Liver Iron Content From Multiecho MR Images.

BACKGROUND: MRI represents the most established liver iron content (LIC) evaluation approach by estimation of liver T2* value, but it is dependent on the choice of the measurement region and the software used for image analysis. PURPOSE: To develop a deep-learning method for unsupervised classification of LIC from magnitude T2* multiecho MR images. STUDY TYPE: Retrospective. POPULATION/SUBJECTS: A total of 1069 thalassemia major patients enrolled in the core laboratory of the Myocardial Iron Overload in Thalassemia (MIOT) network, which were included in the training (80%) and test (20%) sets. Twenty patients from different MRI vendors included in the external test set. FIELD STRENGTH/SEQUENCE: A5 T, T2* multiecho magnitude images. ASSESSMENT: Four deep-learning convolutional neural networks (HippoNet-2D, HippoNet-3D, HippoNet-LSTM, and an ensemble network HippoNet-Ensemble) were used to achieve unsupervised staging of LIC using five classes (normal, borderline, middle, moderate, severe). The training set was employed to construct the deep-learning model. The performance of the LIC staging model was evaluated in the test set and in the external test set. The model's performances were assessed by evaluating the accuracy, sensitivity, and specificity with respect to the ground truth labels obtained by T2* measurements and by comparison with operator-induced variability originating from different region of interest (ROI) placements. STATISTICAL TESTS: The network's performances were evaluated by single-class accuracy, specificity, and sensitivity and compared by one-way repeated measures analysis of variance (ANOVA) and one-way ANOVA. RESULTS: HippoNet-Ensemble reached an accuracy significantly higher than the other networks, and a sensitivity and specificity higher than HippoNet-LSTM. Accuracy, sensitivity, and specificity values for the LIC stages were: normal: 0.96/0.93/0.97, borderline: 0.95/0.85/0.98, mild: 0.96/0.88/0.98, moderate: 0.95/0.89/0.97, severe: 0.97/0.95/0.98. Correctly staging of cases was in the range of 85%-95%, depending on the LIC class. Multiclass accuracy was 0.90 against 0.92 for the interobserver variability. DATA CONCLUSION: The proposed HippoNet-Ensemble network can perform unsupervised LIC staging and achieves good prognostic performance. EVIDENCE LEVEL: 4 TECHNICAL EFFICACY: Stage 2.

Pancreatic fatty replacement as risk marker for altered glucose metabolism and cardiac iron and complications in thalassemia major.

OBJECTIVES: This multicenter study assessed the extent of pancreatic fatty replacement and its correlation with demographics, iron overload, glucose metabolism, and cardiac complications in a cohort of well-treated patients with thalassemia major (TM). METHODS: We considered 308 TM patients (median age: 39.79 years; 182 females) consecutively enrolled in the Extension-Myocardial Iron Overload in Thalassemia Network. Magnetic resonance imaging was used to quantify iron overload (IO) and pancreatic fat fraction (FF) by T2* technique, cardiac function by cine images, and to detect replacement myocardial fibrosis by late gadolinium enhancement technique. The glucose metabolism was assessed by the oral glucose tolerance test. RESULTS: Pancreatic FF was associated with age, body mass index, and history of hepatitis C virus infection. Patients with normal glucose metabolism showed a significantly lower pancreatic FF than patients with impaired fasting glucose (p = 0.030), impaired glucose tolerance (p < 0.0001), and diabetes (p < 0.0001). A normal pancreatic FF (< 6.6%) showed a negative predictive value of 100% for abnormal glucose metabolism. A pancreatic FF > 15.33% predicted the presence of abnormal glucose metabolism. Pancreas FF was inversely correlated with global pancreas and heart T2* values. A normal pancreatic FF showed a negative predictive value of 100% for cardiac iron. Pancreatic FF was significantly higher in patients with myocardial fibrosis (p = 0.002). All patients with cardiac complications had fatty replacement, and they showed a significantly higher pancreatic FF than complications-free patients (p = 0.002). CONCLUSION: Pancreatic FF is a risk marker not only for alterations of glucose metabolism, but also for cardiac iron and complications, further supporting the close link between pancreatic and cardiac disease. KEY POINTS: • In thalassemia major, pancreatic fatty replacement by MRI is a frequent clinical entity, predicted by a pancreas T2* < 20.81 ms and associated with a higher risk of alterations in glucose metabolism. • In thalassemia major, pancreatic fatty replacement is a strong risk marker for cardiac iron, replacement fibrosis, and complications, highlighting a deep connection between pancreatic and cardiac impairment.

Left Ventricle Detection from Cardiac Magnetic Resonance Relaxometry Images Using Visual Transformer.

Left Ventricle (LV) detection from Cardiac Magnetic Resonance (CMR) imaging is a fundamental step, preliminary to myocardium segmentation and characterization. This paper focuses on the application of a Visual Transformer (ViT), a novel neural network architecture, to automatically detect LV from CMR relaxometry sequences. We implemented an object detector based on the ViT model to identify LV from CMR multi-echo T2* sequences. We evaluated performances differentiated by slice location according to the American Heart Association model using 5-fold cross-validation and on an independent dataset of CMR T2*, T2, and T1 acquisitions. To the best of our knowledge, this is the first attempt to localize LV from relaxometry sequences and the first application of ViT for LV detection. We collected an Intersection over Union (IoU) index of 0.68 and a Correct Identification Rate (CIR) of blood pool centroid of 0.99, comparable with other state-of-the-art methods. IoU and CIR values were significantly lower in apical slices. No significant differences in performances were assessed on independent T2* dataset (IoU = 0.68, p = 0.405; CIR = 0.94, p = 0.066). Performances were significantly worse on the T2 and T1 independent datasets (T2: IoU = 0.62, CIR = 0.95; T1: IoU = 0.67, CIR = 0.98), but still encouraging considering the different types of acquisition. This study confirms the feasibility of the application of ViT architectures in LV detection and defines a benchmark for relaxometry imaging.

An Explainable Convolutional Neural Network for the Early Diagnosis of Alzheimer's Disease from 18F-FDG PET.

Convolutional Neural Networks (CNN) which support the diagnosis of Alzheimer's Disease using 18F-FDG PET images are obtaining promising results; however, one of the main challenges in this domain is the fact that these models work as black-box systems. We developed a CNN that performs a multiclass classification task of volumetric 18F-FDG PET images, and we experimented two different post hoc explanation techniques developed in the field of Explainable Artificial Intelligence: Saliency Map (SM) and Layerwise Relevance Propagation (LRP). Finally, we quantitatively analyze the explanations returned and inspect their relationship with the PET signal. We collected 2552 scans from the Alzheimer's Disease Neuroimaging Initiative labeled as Cognitively Normal (CN), Mild Cognitive Impairment (MCI), and Alzheimer's Disease (AD) and we developed and tested a 3D CNN that classifies the 3D PET scans into its final clinical diagnosis. The model developed achieves, to the best of our knowledge, performances comparable with the relevant literature on the test set, with an average Area Under the Curve (AUC) for prediction of CN, MCI, and AD 0.81, 0.63, and 0.77 respectively. We registered the heatmaps with the Talairach Atlas to perform a regional quantitative analysis of the relationship between heatmaps and PET signals. With the quantitative analysis of the post hoc explanation techniques, we observed that LRP maps were more effective in mapping the importance metrics in the anatomic atlas. No clear relationship was found between the heatmap and the PET signal.

Bayesian Convolutional Neural Networks in Medical Imaging Classification: A Promising Solution for Deep Learning Limits in Data Scarcity Scenarios.

Deep neural networks (DNNs) have already impacted the field of medicine in data analysis, classification, and image processing. Unfortunately, their performance is drastically reduced when datasets are scarce in nature (e.g., rare diseases or early-research data). In such scenarios, DNNs display poor capacity for generalization and often lead to highly biased estimates and silent failures. Moreover, deterministic systems cannot provide epistemic uncertainty, a key component to asserting the model's reliability. In this work, we developed a probabilistic system for classification as a framework for addressing the aforementioned criticalities. Specifically, we implemented a Bayesian convolutional neural network (BCNN) for the classification of cardiac amyloidosis (CA) subtypes. We prepared four different CNNs: base-deterministic, dropout-deterministic, dropout-Bayesian, and Bayesian. We then trained them on a dataset of 1107 PET images from 47 CA and control patients (data scarcity scenario). The Bayesian model achieved performances (78.28 (1.99) % test accuracy) comparable to the base-deterministic, dropout-deterministic, and dropout-Bayesian ones, while showing strongly increased "Out of Distribution" input detection (validation-test accuracy mismatch reduction). Additionally, both the dropout-Bayesian and the Bayesian models enriched the classification through confidence estimates, while reducing the criticalities of the dropout-deterministic and base-deterministic approaches. This in turn increased the model's reliability, also providing much needed insights into the network's estimates. The obtained results suggest that a Bayesian CNN can be a promising solution for addressing the challenges posed by data scarcity in medical imaging classification tasks.

Estimation of pancreatic R2* for iron overload assessment in the presence of fat: a comparison of different approaches.

OBJECTIVES: To propose a method for estimating pancreatic relaxation rate, R2*, from conventional multi-echo MRI, based on the nonlinear fitting of the acquired magnitude signal decay to MR signal models that take into account both the signal oscillations induced by fat and the different R2* values of pancreatic parenchyma and fat. MATERIALS AND METHODS: Single-peak fat (SPF) and multi-peak fat (MPF) models were introduced. Single-R2* and dual-R2* assumptions were considered as well. Analyses were conducted on simulated data and 20 thalassemia major patients. RESULTS: Simulations revealed the ability of the MPF model to correctly estimate the R2* value in a large range of fat fractions and R2* values. From the comparison between the results obtained with a single R2* value for water and fat and the dual-R2* approach, the latter is more accurate in both water R2* and fat fraction estimation. In patient's data analysis, a strong concordance was found between SPF and MPF estimated data with measurements done with manual signal correction and from fat-saturated images. The MPF method showed better reproducibility. CONCLUSION: The MPF dual-R2* approach improves reproducibility and reduces image analysis time in the assessment of pancreatic R2* value in patients with iron overload.

Measured PET Data Characterization with the Negative Binomial Distribution Model.

Accurate statistical model of PET measurements is a prerequisite for a correct image reconstruction when using statistical image reconstruction algorithms, or when pre-filtering operations must be performed. Although radioactive decay follows a Poisson distribution, deviation from Poisson statistics occurs on projection data prior to reconstruction due to physical effects, measurement errors, correction of scatter and random coincidences. Modelling projection data can aid in understanding the statistical nature of the data in order to develop efficient processing methods and to reduce noise. This paper outlines the statistical behaviour of measured emission data evaluating the goodness of fit of the negative binomial (NB) distribution model to PET data for a wide range of emission activity values. An NB distribution model is characterized by the mean of the data and the dispersion parameter α that describes the deviation from Poisson statistics. Monte Carlo simulations were performed to evaluate: (a) the performances of the dispersion parameter α estimator, (b) the goodness of fit of the NB model for a wide range of activity values. We focused on the effect produced by correction for random and scatter events in the projection (sinogram) domain, due to their importance in quantitative analysis of PET data. The analysis developed herein allowed us to assess the accuracy of the NB distribution model to fit corrected sinogram data, and to evaluate the sensitivity of the dispersion parameter α to quantify deviation from Poisson statistics. By the sinogram ROI-based analysis, it was demonstrated that deviation on the measured data from Poisson statistics can be quantitatively characterized by the dispersion parameter α, in any noise conditions and corrections.

Variable density randomized stack of spirals (VDR-SoS) for compressive sensing MRI.

PURPOSE: To develop a 3D sampling strategy based on a stack of variable density spirals for compressive sensing MRI. METHODS: A random sampling pattern was obtained by rotating each spiral by a random angle and by delaying for few time steps the gradient waveforms of the different interleaves. A three-dimensional (3D) variable sampling density was obtained by designing different variable density spirals for each slice encoding. The proposed approach was tested with phantom simulations up to a five-fold undersampling factor. Fully sampled 3D dataset of a human knee, and of a human brain, were obtained from a healthy volunteer. The proposed approach was tested with off-line reconstructions of the knee dataset up to a four-fold acceleration and compared with other noncoherent trajectories. RESULTS: The proposed approach outperformed the standard stack of spirals for various undersampling factors. The level of coherence and the reconstruction quality of the proposed approach were similar to those of other trajectories that, however, require 3D gridding for the reconstruction. CONCLUSION: The variable density randomized stack of spirals (VDR-SoS) is an easily implementable trajectory that could represent a valid sampling strategy for 3D compressive sensing MRI. It guarantees low levels of coherence without requiring 3D gridding. Magn Reson Med 76:59-69, 2016. © 2015 Wiley Periodicals, Inc.

MRI Application and Challenges of Hyperpolarized Carbon-13 Pyruvate in Translational and Clinical Cardiovascular Studies: A Literature Review.

Cardiovascular disease shows, or may even be caused by, changes in metabolism. Hyperpolarized magnetic resonance spectroscopy and imaging is a technique that could assess the role of different aspects of metabolism in heart disease, allowing real-time metabolic flux assessment in vivo. In this review, we introduce the main hyperpolarization techniques. Then, we summarize the use of dedicated radiofrequency 13C coils, and report a state of the art of 13C data acquisition. Finally, this review provides an overview of the pre-clinical and clinical studies on cardiac metabolism in the healthy and diseased heart. We furthermore show what advances have been made to translate this technique into the clinic in the near future and what technical challenges still remain, such as exploring other metabolic substrates.

Automatic 2D registration of renal perfusion image sequences by mutual information and adaptive prediction.

UNLABELLED: The objective of this study was to develop an automatic image registration technique capable of compensating for kidney motion in renal perfusion MRI, to assess the effect of renal artery stenosis on the kidney parenchyma. MATERIALS AND METHODS: Images from 20 patients scheduled for a renal perfusion study were acquired using a 1.5 T scanner. A free-breathing 3D-FSPGR sequence was used to acquire coronal views encompassing both kidneys following the infusion of Gd-BOPTA. A two-step registration algorithm was developed, including a preliminary registration minimising the quadratic difference and a fine registration maximising the mutual information (MI) between consecutive image frames. The starting point for the MI-based registration procedure was provided by an adaptive predictor that was able to predict kidney motion using a respiratory movement model. The algorithm was validated against manual registration performed by an expert user. RESULTS: The mean distance between the automatically and manually defined contours was 2.95 ± 0.81 mm, which was not significantly different from the interobserver variability of the manual registration procedure (2.86 ± 0.80 mm, P = 0.80). The perfusion indices evaluated on the manually and automatically extracted perfusion curves were not significantly different. CONCLUSIONS: The developed method is able to automatically compensate for kidney motion in perfusion studies, which prevents the need for time-consuming manual image registration.

How the signal-to-noise ratio influences hyperpolarized 13C dynamic MRS data fitting and parameter estimation.

MRS of hyperpolarized (13) C-labeled compounds represents a promising technique for in vivo metabolic studies. However, robust quantification and metabolic modeling are still important areas of investigation. In particular, time and spatial resolution constraints may lead to the analysis of MRS signals with low signal-to-noise ratio (SNR). The relationship between SNR and the precision of quantitative analysis for the evaluation of the in vivo kinetic behavior of metabolites is unknown. In this article, this topic is addressed by Monte Carlo simulations, covering the problem of MRS signal model parameter estimation, with strong emphasis on the peak amplitude and kinetic model parameters. The results of Monte Carlo simulation were confirmed by in vivo experiments on medium-sized animals injected with hyperpolarized [1-(13) C]pyruvate. The results of this study may be useful for the establishment of experimental planning and for the optimization of kinetic model estimation as a function of the SNR value.

A novel tool for estimation of magnetic resonance occupational exposure to spatially varying magnetic fields.

OBJECT: Staff operating in the environment of magnetic resonance imaging (MRI) scanners are exposed daily to static magnetic fields (MFs). To protect workers several guidelines are present in literature reporting exposure limits values expressed in terms of magnetic flux density or induced current density. We present here a novel tool for estimating the induced current density due to worker movement in the MR environment. MATERIALS AND METHODS: A Matlab script was created to estimate the induced current density J due to operator movements along a chosen walking path. RESULTS: The induced current density associated with any worker's movements during MR procedures is dependent on the walking speed and on the spatial gradient fields associated with a specific path. Some examples of possible worker paths were considered here for a 3 T MR scanner and a maximum value of 160 cm/s walking speed. CONCLUSION: This tool permits one to find exposure level for specific worker walking path and speed; it can be used as assessment tool in any MRI centre and for workers safety education. It is valid for any kind of commercial scanner because it requires only the knowledge of the MR scanner room map with isogauss lines.

Deep-learning-based cardiac amyloidosis classification from early acquired pet images.

The objective of the present work was to evaluate the potential of deep learning tools for characterizing the presence of cardiac amyloidosis from early acquired PET images, i.e. 15 min after [18F]-Florbetaben tracer injection. 47 subjects were included in the study: 13 patients with transthyretin-related amyloidosis cardiac amyloidosis (ATTR-CA), 15 patients with immunoglobulin light-chain amyloidosis (AL-CA), and 19 control-patients (CTRL). [18F]-Florbetaben PET/CT images were acquired in list mode and data was sorted into a sinogram, covering a time interval of 5 min starting 15 min after the injection. The resulting sinogram was reconstructed using OSEM iterative algorithm. A deep convolutional neural network (CAclassNet) was designed and implemented, consisting of five 2D convolutional layers, three fully connected layers and a final classifier returning AL, ATTR and CTRL scores. A total of 1107 2D images (375 from AL-subtype patients, 312 from ATTR-subtype, and 420 from Controls) have been considered in the study and used to train, validate and test the proposed network. CAclassNet cross-validation resulted with train error mean ± sd of 2.001% ± 0.96%, validation error of 4.5% ± 2.26%, and net accuracy of 95.49% ± 2.26%. Network test error resulted in a mean ± sd values of 10.73% ± 0.76%. Sensitivity, specificity, and accuracy evaluated on the test dataset were respectively for AL-CA sub-type: 1, 0.912, 0.936; for ATTR-CA: 0.935, 0.897, 0.972; for control subjects: 0.809, 0.971, 0.909. In conclusion, the proposed CAclassNet model seems very promising as an aid for the clinician in the diagnosis of CA from cardiac [18F]-Florbetaben PET images acquired a few minutes after the injection.

[18F]-Florbetaben PET/CT for Differential Diagnosis Among Cardiac Immunoglobulin Light Chain, Transthyretin Amyloidosis, and Mimicking Conditions.

OBJECTIVES: This study aimed to test the diagnostic value of [18F]-florbetaben positron emission tomography (PET) in patients with suspicion of CA. BACKGROUND: Diagnosis of cardiac involvement in immunoglobulin light-chain-derived amyloidosis (AL) and transthyretin-related amyloidosis (ATTR), which holds major importance in risk stratification and decision making, is frequently delayed. Furthermore, although diphosphonate radiotracers allow a noninvasive diagnosis of ATTR, demonstration of cardiac amyloidosis (CA) in AL may require endomyocardial biopsy. METHODS: Forty patients with biopsy-proven diagnoses of CA (20 ALs, 20 ATTRs) and 20 patients referred with the initial clinical suspicion and later diagnosed with non-CA pathology underwent a cardiac PET/computed tomography scan with a 60-min dynamic [18F]-florbetaben PET acquisition, and 4 10-min static scans at 5, 30, 50, and 110 min after radiotracer injection. RESULTS: Visual qualitative assessment showed intense early cardiac uptake in all subsets. Patients with AL displayed a high, persistent cardiac uptake in all the static scans, whereas patients with ATTR and those with non-CA showed an uptake decrease soon after the early scan. Semiquantitative assessment demonstrated higher mean standardized uptake value (SUVmean) in patients with AL, sustained over the whole acquisition period (early SUVmean: 5.55; interquartile range [IQR]: 4.00 to 7.43; vs. delayed SUVmean: 3.50; IQR: 2.32 to 6.10; p = NS) compared with in patients with ATTR (early SUVmean: 2.55; IQR: 1.80 to 2.97; vs. delayed SUVmean: 1.25; IQR: 0.90 to 1.60; p < 0.001) and in patients with non-CA (early SUVmean: 3.50; IQR: 1.60 to 3.37; vs. delayed SUVmean: 1.40; IQR: 1.20 to 1.60; p < 0.001). Similar results were found comparing heart-to-background ratio and molecular volume. CONCLUSIONS: Delayed [18F]-florbetaben cardiac uptake may discriminate CA due to AL from either ATTR or other mimicking conditions. [18F]-florbetaben PET/computed tomography may represent a promising noninvasive tool for the diagnosis of AL amyloidosis, which is still often challenging and delayed. (A Prospective Triple-Arm, Monocentric, Phase-II Explorative Study on Evaluation of Diagnostic Efficacy of the PET Tracer [18F]-Florbetaben [Neuraceq] in Patients With Cardiac Amyloidosis [FLORAMICAR2]; EudraCT number: 2017-001660-38).

Preferential patterns of myocardial iron overload by multislice multiecho T*2 CMR in thalassemia major patients.

T*(2) multislice multiecho cardiac MR allows quantification of the segmental distribution of myocardial iron overload. This study aimed to determine if there were preferential patterns of myocardial iron overload in thalassemia major. Five hundred twenty-three thalassemia major patients underwent cardiac MR. Three short-axis views of the left ventricle were acquired and analyzed using a 16-segment standardized model. The T*(2) value on each segment was calculated, as well as the global value. Four main circumferential regions (anterior, septal, inferior, and lateral) were defined. Significant segmental variability was found in the 229 patients with significant myocardial iron overload (global T*(2) <26 ms), subsequently divided into two groups: severe (global T*(2) <10 ms) and mild to moderate (global T*(2) between 10 and 26 ms) myocardial iron overload. A preferential pattern of iron store in anterior and inferior regions was detected in both groups. This pattern was preserved among the slices. The pattern could not be explained by additive susceptibility artifacts, negligible in heavily iron-loaded patients. A significantly higher T*(2) value in the basal slice was found in patients with severe iron overload. In conclusion, a segmental T*(2) cardiac MR approach could identify early iron deposit, useful for tailoring chelation therapy and preventing myocardial dysfunction in the clinical setting.

Hyperpolarized 13C MRS surface coil: design and signal-to-noise ratio estimation.

PURPOSE: Hyperpolarized carbon-13 magnetic resonance spectroscopy is a novel and powerful tool for exploring the metabolic state of tissue, but a number of technological problems still limit this technology and need innovative solutions. In particular, the low molar concentration of derivate metabolites give rise to low signal-to-noise ratio (SNR), which makes the design and development of dedicated RF coils a task of fundamental importance. In this article, the authors describe the simulation and the design of a dedicated 13C surface coil for cardiac metabolism assessment in pig models. METHODS: A SNR model for a circular loop is presented and applied to the design of a 13C coil which guarantees the desired field-of-view and provides high SNR with a good penetration in deep sample regions. The coil resistance was calculated from Ohm's law and the magnetic field pattern was calculated using Biot-Savart law, while the sample induced resistance was calculated using a numerical finite-difference time-domain algorithm. Successively, a prototype of the coil was built and tested on the workbench and by acquisition of MR data. RESULTS: The comparison of SNR-vs-depth profiles between the theoretical SNR model and the experimental SNR extracted from the phantom chemical shift image (CSI) showed the accuracy of the authors' model. Moreover, the authors demonstrated the use of the coil for the acquisition of a CSI of a hyperpolarized [1-13C] pyruvate phantom. CONCLUSIONS: The results demonstrated the design trade-offs to successfully design a dedicated coil for cardiac imaging in the pig with hyperpolarized 13C by developing a SNR model which allows the prediction of the coil performance. This approach can be employed for deriving SNR formulations for coil with more complex geometries.

Probabilistic Graphical Models for Dynamic PET: A Novel Approach to Direct Parametric Map Estimation and Image Reconstruction.

In the context of dynamic emission tomography, the conventional processing pipeline consists of independent image reconstruction of single-time frames, followed by the application of a suitable kinetic model to time-activity curves (TACs) at the voxel or region-of-interest level. Direct 4D positron emission tomography (PET) reconstruction, by contrast, seeks to move beyond this scheme and incorporate information from multiple time frames within the reconstruction task. Established direct methods are based on a deterministic description of voxelwise TACs, captured by the chosen kinetic model, considering the photon counting process the only source of uncertainty. In this paper, we introduce a new probabilistic modeling strategy based on the key assumption that activity time course would be subject to uncertainty even if the parameters of the underlying dynamic process are known. This leads to a hierarchical model that we formulate using the formalism of probabilistic graphical modeling. The inference is addressed using a new iterative algorithm, in which kinetic modeling results are treated as prior expectation of activity time course, rather than as a deterministic match, making it possible to control the trade-off between a data-driven and a model-driven reconstruction. The proposed method is flexible to an arbitrary choice of (linear and nonlinear) kinetic models, it enables the inclusion of arbitrary (sub)differentiable priors for parametric maps, and it is simple to implement. Computer simulations and an application to a real-patient scan show how the proposed method is able to generalize over conventional indirect and direct approaches, providing a bridge between them by properly tuning the impact of the kinetic modeling step on image reconstruction.

Multislice multiecho T2* cardiac magnetic resonance for the detection of heterogeneous myocardial iron distribution in thalassaemia patients.

The present study investigated myocardial T2* heterogeneity in thalassaemia major (TM) patients by cardiac magnetic resonance (CMR), to determine whether is related to inhomogeneous iron overload distribution. A total of 230 TM patients consecutively referred to our laboratory were studied retrospectively. Three short-axis views (basal, medium and apical) of the left ventricle (LV) were obtained by multislice multiecho T2* CMR. T2* segmental distribution was mapped on a 16-segment LV model. The level of heterogeneity of the T2* segmental distribution, evaluated by the coefficient of variation (CoV), was compared with that of a surrogate data set, to determine whether the inhomogeneous segmental distribution of T2* could be generated by susceptibility artefacts. Susceptibility artefacts offer an explanation for the T2* heterogeneity observed in patients without iron overload. In subjects with global T2* below the lower limit of the normal, T2* heterogeneity increased abruptly which could not be explained by artefactual effects. Some segmental T2* values were below and others above the limit of normal threshold (20 ms) in 104 (45%) TM patients. Among these patients, 74% showed a normal T2* global value. In conclusion, a true heterogeneity in the iron overload distribution may be present in TM patients. Heterogeneity seemingly appears in the borderline myocardial iron and stabilizes at moderate to severe iron burden.

Improved T2* assessment in liver iron overload by magnetic resonance imaging.

In the clinical MRI practice, it is common to assess liver iron overload by T2* multi-echo gradient-echo images. However, there is no full consensus about the best image analysis approach for the T2* measurements. The currently used methods involve manual drawing of a region of interest (ROI) within MR images of the liver. Evaluation of a representative liver T2* value is done by fitting an appropriate model to the signal decay within the ROIs vs. the echo time. The resulting T2* value may depend on both ROI placement and choice of the signal decay model. The aim of this study was to understand how the choice of the analysis methodology may affect the accuracy of T2* measurements. A software model of the iron overloaded liver was inferred from MR images acquired from 40 thalassemia major patients. Different image analysis methods were compared exploiting the developed software model. Moreover, a method for global semiautomatic T2* measurement involving the whole liver was developed. The global method included automatic segmentation of parenchyma by an adaptive fuzzy-clustering algorithm able to compensate for signal inhomogeneities. Global liver T2* value was evaluated using a pixel-wise technique and an optimized signal decay model. The global approach was compared with the ROI-based approach used in the clinical practice. For the ROI-based approach, the intra-observer and inter-observer coefficients of variation (CoVs) were 3.7% and 5.6%, respectively. For the global analysis, the CoVs for intra-observers and inter-observers reproducibility were 0.85% and 2.87%, respectively. The variability shown by the ROI-based approach was acceptable for use in the clinical practice; however, the developed global method increased the accuracy in T2* assessment and significantly reduced the operator dependence and sampling errors. This global approach could be useful in the clinical arena for patients with borderline liver iron overload and/or requiring follow-up studies.