Unsupervised deep learning model for correcting Nyquist ghosts of single-shot spatiotemporal encoding.

PURPOSE: To design an unsupervised deep learning (DL) model for correcting Nyquist ghosts of single-shot spatiotemporal encoding (SPEN) and evaluate the model for real MRI applications. METHODS: The proposed method consists of three main components: (1) an unsupervised network that combines Residual Encoder and Restricted Subspace Mapping (RERSM-net) and is trained to generate a phase-difference map based on the even and odd SPEN images; (2) a spin physical forward model to obtain the corrected image with the learned phase difference map; and (3) cycle-consistency loss that is explored for training the RERSM-net. RESULTS: The proposed RERSM-net could effectively generate smooth phase difference maps and correct Nyquist ghosts of single-shot SPEN. Both simulation and real in vivo MRI experiments demonstrated that our method outperforms the state-of-the-art SPEN Nyquist ghost correction method. Furthermore, the ablation experiments of generating phase-difference maps show the advantages of the proposed unsupervised model. CONCLUSION: The proposed method can effectively correct Nyquist ghosts for the single-shot SPEN sequence.

Quantifying H&E staining results, grading and predicting IDH mutation status of gliomas using hybrid multi-dimensional MRI.

OBJECTIVE: To assess the performance of hybrid multi-dimensional magnetic resonance imaging (HM-MRI) in quantifying hematoxylin and eosin (H&E) staining results, grading and predicting isocitrate dehydrogenase (IDH) mutation status of gliomas. MATERIALS AND METHODS: Included were 71 glioma patients (mean age, 50.17 ± 13.38 years; 35 men). HM-MRI images were collected at five different echo times (80-200 ms) with seven b-values (0-3000 s/mm2). A modified three-compartment model with very-slow, slow and fast diffusion components was applied to calculate HM-MRI metrics, including fractions, diffusion coefficients and T2 values of each component. Pearson correlation analysis was performed between HM-MRI derived fractions and H&E staining derived percentages. HM-MRI metrics were compared between high-grade and low-grade gliomas, and between IDH-wild and IDH-mutant gliomas. Using receiver operational characteristic (ROC) analysis, the diagnostic performance of HM-MRI in grading and genotyping was compared with mono-exponential models. RESULTS: HM-MRI metrics FDvery-slow and FDslow demonstrated a significant correlation with the H&E staining results (p < .05). Besides, FDvery-slow showed the highest area under ROC curve (AUC = 0.854) for grading, while Dslow showed the highest AUC (0.845) for genotyping. Furthermore, a combination of HM-MRI metrics FDvery-slow and T2Dslow improved the diagnostic performance for grading (AUC = 0.876). DISCUSSION: HM-MRI can aid in non-invasive diagnosis of gliomas.

Differentiation of malignant from benign pleural effusions based on artificial intelligence.

INTRODUCTION: This study aimed to construct artificial intelligence models based on thoracic CT images to perform segmentation and classification of benign pleural effusion (BPE) and malignant pleural effusion (MPE). METHODS: A total of 918 patients with pleural effusion were initially included, with 607 randomly selected cases used as the training cohort and the other 311 as the internal testing cohort; another independent external testing cohort with 362 cases was used. We developed a pleural effusion segmentation model (M1) by combining 3D spatially weighted U-Net with 2D classical U-Net. Then, a classification model (M2) was built to identify BPE and MPE using a CT volume and its 3D pleural effusion mask as inputs. RESULTS: The average Dice similarity coefficient, Jaccard coefficient, precision, sensitivity, Hausdorff distance 95% (HD95) and average surface distance indicators in M1 were 87.6±5.0%, 82.2±6.2%, 99.0±1.0%, 83.0±6.6%, 6.9±3.8 and 1.6±1.1, respectively, which were better than those of the 3D U-Net and 3D spatially weighted U-Net. Regarding M2, the area under the receiver operating characteristic curve, sensitivity and specificity obtained with volume concat masks as input were 0.842 (95% CI 0.801 to 0.878), 89.4% (95% CI 84.4% to 93.2%) and 65.1% (95% CI 57.3% to 72.3%) in the external testing cohort. These performance metrics were significantly improved compared with those for the other input patterns. CONCLUSIONS: We applied a deep learning model to the segmentation of pleural effusions, and the model showed encouraging performance in the differential diagnosis of BPE and MPE.

Unsupervised cycle-consistent network using restricted subspace field map for removing susceptibility artifacts in EPI.

PURPOSE: To design an unsupervised deep neural model for correcting susceptibility artifacts in single-shot Echo Planar Imaging (EPI) and evaluate the model for preclinical and clinical applications. METHODS: This work proposes an unsupervised cycle-consistent model based on the restricted subspace field map to take advantage of both the deep learning (DL) and the reverse polarity-gradient (RPG) method for single-shot EPI. The proposed model consists of three main components: (1) DLRPG neural network (DLRPG-net) to obtain field maps based on a pair of images acquired with reversed phase encoding; (2) spin physical model-based modules to obtain the corrected undistorted images based on the learned field map; and (3) cycle-consistency loss between the input images and back-calculated images from each cycle is explored for network training. In addition, the field maps generated by DLRPG-net belong to a restricted subspace, which is a span of predefined cubic splines to ensure the smoothness of the field maps and avoid blurring in the corrected images. This new method is trained and validated on both preclinical and clinical datasets for diffusion MRI. RESULTS: The proposed network could effectively generate smooth field maps and correct susceptibility artifacts in single-shot EPI. Simulated and in vivo preclinical/clinical experiments demonstrated that our method outperforms the state-of-the-art susceptibility artifact correction methods. Furthermore, the ablation experiments of the cycle-consistent network and the restricted subspace in generating field maps did show the advantages of DLRPG-net. CONCLUSION: The proposed method (DLRPG-net) can effectively correct susceptibility artifacts for preclinical and clinical single-shot EPI sequences.

Deuterium imaging of the Warburg effect at sub-millimolar concentrations by joint processing of the kinetic and spectral dimensions.

Deuterium metabolic imaging (DMI) is a promising molecular MRI approach, which follows the administration of deuterated substrates and their metabolization. [6,6'-2 H2 ]-glucose for instance is preferentially converted in tumors to [3,3'-2 H2 ]-lactate as a result of the Warburg effect, providing a distinct resonance whose mapping using time-resolved spectroscopic imaging can diagnose cancer. The MR detection of low-concentration metabolites such as lactate, however, is challenging. It has been recently shown that multi-echo balanced steady-state free precession (ME-bSSFP) increases the signal-to-noise ratio (SNR) of these experiments approximately threefold over regular chemical shift imaging; the present study examines how DMI's sensitivity can be increased further by advanced processing methods. Some of these, such as compressed sensing multiplicative denoising and block-matching/3D filtering, can be applied to any spectroscopic/imaging methods. Sensitivity-enhancing approaches were also specifically tailored to ME-bSSFP DMI, by relying on priors related to the resonances' positions and to features of the metabolic kinetics. Two new methods are thus proposed that use these constraints for enhancing the sensitivity of both the spectral images and the metabolic kinetics. The ability of these methods to improve DMI is evidenced in pancreatic cancer studies carried at 15.2 T, where suitable implementations of the proposals imparted eightfold or more SNR improvement over the original ME-bSSFP data, at no informational cost. Comparisons with other propositions in the literature are briefly discussed.

Multimodal Imaging Response after the Singular or Combination Treatments of Vascular Endothelial Growth Factor Inhibitor and Immune Checkpoint Inhibitor.

This study aims to dynamically assess tumor changes after variable treatments with vascular endothelial growth factor (VEGF) inhibitor and/or immune checkpoint inhibitor (ICI) using multimodal imaging of MRI and 18F-FDG PET/CT in a hepatocellular carcinoma (HCC) mice model. Based on different treatments, 24 mice were randomly divided into four groups: control (isotype-matched IgG antibody 10 mg/kg), VEGF inhibitor (sorafenib 50 mg/kg), ICI (anti-PD-L1 antibody 10 mg/kg), and combination groups (sorafenib 50 mg/kg + anti-PD-L1 antibody 10 mg/kg). Quantitative imaging assessments, including volume transfer constant (Ktrans), apparent diffusion coefficient (ADC), lactate/choline ratio, and the maximum standardized 18F-FDG uptake value ratio of tumor to muscle (SUVtumor/SUVmuscle ratio), were acquired at different time points (before treatment and 7, 14, and 21 days after treatment). Quantitative data were presented as the mean ± standard errors and two-way repeated-measure ANOVA tests were performed for intergroup and intertime point comparisons. After 21 days from the initiation of therapies, combination group showed the lowest tumor volume and weight, followed by ICI, VEGF inhibitor, and control group, with no significance between the VEGF inhibitor and control groups. In addition, Ktrans values significantly decreased, and the lactate/choline ratio and SUVtumor/SUVmuscle ratio were significantly elevated in the VEGF inhibitor group. ADC significantly increased in the ICI and combination groups, with no significant differences in ADC observed between the control and VEGF inhibitor groups, which showed a similar dynamic change to the tumor volume. Furthermore, Ktrans, lactate/choline ratio, and ADC were significantly correlated with CD31+ area, hypoxyprobe+ area, and apoptosis, respectively. Our results suggest that the singular treatment and combination of the VEGF inhibitor and ICI treatments for HCC present different multimodal imaging changes in accordance with the specific histopathological features. These findings might facilitate the formulation of better treatment response criteria; besides, we find ADC is probably an indicator easily to obtain for treatment response evaluation.

Improving deuterium metabolic imaging (DMI) signal-to-noise ratio by spectroscopic multi-echo bSSFP: A pancreatic cancer investigation.

PURPOSE: Deuterium metabolic imaging (DMI) maps the uptake of deuterated precursors and their conversion into lactate and other markers of tumor metabolism. Even after leveraging 2 H's short T1 s, DMI's signal-to-noise ratio (SNR) is limited. We hypothesize that a multi-echo balanced steady-state free precession (ME-bSSFP) approach would increase SNR compared to chemical shift imaging (CSI), while achieving spectral isolation of the metabolic precursors and products. METHODS: Suitably tuned 2 H ME-bSSFP (five echo times [TEs], ΔTE = 2.2 ms, repetition time [TR]/flip-angle = 12 ms/60°) was implemented at 15.2T and compared to CSI (TR/flip-angle = 95 ms/90°) regarding SNR and spectral isolation, in simulations, in deuterated phantoms and for the in vivo diagnosis of a mouse tumor model of pancreatic adenocarcinoma (N = 10). RESULTS: Simulations predicted an SNR increase vs. CSI of 3-5, and that the peaks of 2 H-water, 2 H6,6' -glucose, and 2 H3,3' -lactate can be well isolated by ME-bSSFP; phantoms confirmed this. In vivo, at equal spatial resolution (1.25 × 1.25 mm2 ) and scan time (10 min), 2 H6,6' -glucose's and 2 H3,3' -lactate's SNR were indeed higher for bSSFP than for CSI, three-fold for glucose (57 ± 30 vs. 19 ± 11, P < .001), doubled for water (13 ± 5 vs. 7 ± 3, P = .005). The time courses and overall localization of all metabolites agreed well, comparing CSI against ME-bSSFP. However, a clearer localization of glucose in kidneys and bladder, the detection of glucose-avid rims in certain tumors, and a heterogeneous pattern of intra-tumor lactate production could only be observed using ME-bSSFP's higher resolution. CONCLUSIONS: ME-bSSFP provides greater SNR per unit time than CSI, providing for higher spatial resolution mapping of glucose uptake and lactate production in tumors.

Identification of variable stages in murine pancreatic tumors by a multiparametric approach employing hyperpolarized 13 C MRSI, 1 H diffusivity and 1 H T1 MRI.

This study explored the usefulness of multiple quantitative MRI approaches to detect pancreatic ductal adenocarcinomas in two murine models, PAN-02 and KPC. Methods assayed included 1 H T1 and T2 measurements, quantitative diffusivity mapping, magnetization transfer (MT) 1 H MRI throughout the abdomen and hyperpolarized 13 C spectroscopic imaging. The progress of the disease was followed as a function of its development; studies were also conducted for wildtype control mice and for mice with induced mild acute pancreatitis. Customized methods developed for scanning the motion- and artifact-prone mice abdomens allowed us to obtain quality 1 H images for these targeted regions. Contrasts between tumors and surrounding tissues, however, were significantly different. Anatomical images, T2 maps and MT did not yield significant contrast unless tumors were large. By contrast, tumors showed statistically lower diffusivities than their surroundings (≈8.3 ± 0.4 x 10-4 for PAN-02 and ≈10.2 ± 0.6 x 10-4 for KPC vs 13 ± 1 x 10-3 mm2 s-1 for surroundings), longer T1 relaxation times (≈1.44 ± 0.05 for PAN-02 and ≈1.45 ± 0.05 for KPC vs 0.95 ± 0.10 seconds for surroundings) and significantly higher lactate/pyruvate ratios by hyperpolarized 13 C MR (0.53 ± 0.2 for PAN-02 and 0.78 ± 0.2 for KPC vs 0.11 ± 0.04 for control and 0.31 ± 0.04 for pancreatitis-bearing mice). Although the latter could also distinguish early-stage tumors from healthy animal controls, their response was similar to that in our pancreatitis model. Still, this ambiguity could be lifted using the 1 H-based reporters. If confirmed for other kinds of pancreatic tumors this means that these approaches, combined, can provide a route to an early detection of pancreatic cancer.

Dynamic T2 mapping by multi-spin-echo spatiotemporal encoding.

PURPOSE: To develop a pulse sequence for acquiring robust, quantitative T2 relaxation maps in real time. METHODS: The pulse scheme relies on fully refocused spatiotemporally encoded multi-spin-echo trains, which provide images that are significantly less distorted than spin-echo echo planar imaging-based counterparts. This enables single-shot T2 mapping in inhomogeneity-prone regions. Another advantage of these schemes stems from their ability to interleave multiple scans in a reference-free manner, providing an option to increase sensitivity and spatial resolution with minimal motional artifacts. RESULTS: The method was implemented in preclinical and clinical scanners, where single-shot acquisitions delivered reliable T2 maps in ≤200 ms with ≈250 µm and ≈3 mm resolutions, respectively. Ca. 4 times higher spatial resolutions were achieved for the motion-compensated interleaved versions of these acquisitions, delivering T2 maps in ca. 10 s per slice. These maps were nearly indistinguishable from multi-scan relaxometric maps requiring orders-of-magnitude longer acquisitions; this was confirmed by mice head and real-time mice abdomen 7T scans performed following contrast-agent injections, as well as by 3T human brain and breast scans. CONCLUSION: This study introduced and demonstrated a new approach for acquiring rapid and quantitative T2 data, which is particularly reliable when operating at high fields and/or targeting heterogeneous organs or regions.

High-resolution diffusion MRI studies of development in pregnant mice visualized by novel spatiotemporal encoding schemes.

This study introduces an MRI approach to map diffusion of water in vivo with high resolution under challenging conditions; the approach's potential is then used in diffusivity characterizations of embryos and fetoplacental units in pregnant mice, as well as of newborn mice in their initial postnatal period. The method relies on performing self-referenced spatiotemporal encoded MRI acquisitions, which can achieve the motional and susceptibility immunities needed to target challenging regions such as a mouse's abdominal cavity in a single shot. When suitably combined with zooming-in and novel interleaving procedures, these scans can overcome the inhomogeneity and sensitivity challenges arising upon targeting ≈100 µm in-plane resolutions, and thereby enable longitudinal development studies of abdominal organs that have hitherto eluded in vivo diffusion-weighted imaging. This is employed here to follow processes related to embryonic implantation and placentation, including the final stages of mouse gastrulation, the development of white matter in fetal brains, the maturation of fetal spines, and the evolution of the different layers making up mouse hemochorial placentas. The protocol's ability to extract diffusivity information in challenging regions as a function of embryonic mouse development is thus demonstrated, and its usefulness as a tool for visualizing pregnancy-related developmental changes in rodents is discussed.

Diffusion tensor distribution imaging of an in vivo mouse brain at ultrahigh magnetic field by spatiotemporal encoding.

Diffusion tensor distribution (DTD) imaging builds on principles from diffusion, solid-state and low-field NMR spectroscopies, to quantify the contents of heterogeneous voxels as nonparametric distributions, with tensor "size", "shape" and orientation having direct relations to corresponding microstructural properties of biological tissues. The approach requires the acquisition of multiple images as a function of the magnitude, shape and direction of the diffusion-encoding gradients, leading to long acquisition times unless fast image read-out techniques like EPI are employed. While in previous in vivo human brain studies performed at 3 T this proved a viable option, porting these measurements to very high magnetic fields and/or to heterogeneous organs induces B0 - and B1 -inhomogeneity artifacts that challenge the limits of EPI. To overcome such challenges, we demonstrate here that high spatial resolution DTD of mouse brain can be carried out at 15.2 T with a surface-cryoprobe, by relying on SPatiotemporal ENcoding (SPEN) imaging sequences. These new acquisition and data-processing protocols are demonstrated with measurements on in vivo mouse brain, and validated with synthetic phantoms designed to mimic the diffusion properties of white matter, gray matter and cerebrospinal fluid. While still in need of full extensions to 3D mappings and of scanning additional animals to extract more general physiological conclusions, this work represents another step towards the model-free, noninvasive in vivo characterization of tissue microstructure and heterogeneity in animal models, at ≈0.1 mm resolutions.

Stationary wavelet denoising of solid-state NMR spectra using multiple similar measurements.

Accumulating several scans of free induction decays is always needed to improve the signal-to-noise ratio of NMR spectra, especially for the low gyromagnetic ratio solid-state NMR. In this study, we present a new denoising approach based on the correlations between multiple similar NMR spectra. Contrary to the simple averaging of multiple scans or denoising the final averaged spectrum, we propose a Wavelet-based Denoising technique for Multiple Similar scans(WDMS). Firstly, the stationary wavelet transform is applied to decompose every spectrum into approximation coefficients and detail coefficients. Then, the detail coefficients are multiplied by weights calculated based on Pearson's correlation coefficient and structural similarity index between approximation coefficients of different spectra. Finally, the average of these detailed components is used to denoise the spectra. The proposed method is carried on the assumption that noise between multiple spectra is uncorrelated while peak signal information is similar between different spectra, thus preserving the possibility of applying further processing to the data. As a demonstration, the standard wavelet denoise is applied to the WDMS-processed spectra, achieving a further increase in the S/N ratio. We confirm the reliability of the denoising approach based on multiple scans on 1D/2D solid-state MAS/static NMR spectra. In addition, we also show that this method can be used to deal with a single Car-Purcell-Meiboom-Gill (CPMG) echo train.

Advances and prospects in deuterium metabolic imaging (DMI): a systematic review of in vivo studies.

BACKGROUND: Deuterium metabolic imaging (DMI) has emerged as a promising non-invasive technique for studying metabolism in vivo. This review aims to summarize the current developments and discuss the futures in DMI technique in vivo. METHODS: A systematic literature review was conducted based on the PRISMA 2020 statement by two authors. Specific technical details and potential applications of DMI in vivo were summarized, including strategies of deuterated metabolites detection, deuterium-labeled tracers and corresponding metabolic pathways in vivo, potential clinical applications, routes of tracer administration, quantitative evaluations of metabolisms, and spatial resolution. RESULTS: Of the 2,248 articles initially retrieved, 34 were finally included, highlighting 2 strategies for detecting deuterated metabolites: direct and indirect DMI. Various deuterated tracers (e.g., [6,6'-2H2]glucose, [2,2,2'-2H3]acetate) were utilized in DMI to detect and quantify different metabolic pathways such as glycolysis, tricarboxylic acid cycle, and fatty acid oxidation. The quantifications (e.g., lactate level, lactate/glutamine and glutamate ratio) hold promise for diagnosing malignancies and assessing early anti-tumor treatment responses. Tracers can be administered orally, intravenously, or intraperitoneally, either through bolus administration or continuous infusion. For metabolic quantification, both serial time point methods (including kinetic analysis and calculation of area under the curves) and single time point quantifications are viable. However, insufficient spatial resolution remains a major challenge in DMI (e.g., 3.3-mL spatial resolution with 10-min acquisition at 3 T). CONCLUSIONS: Enhancing spatial resolution can facilitate the clinical translation of DMI. Furthermore, optimizing tracer synthesis, administration protocols, and quantification methodologies will further enhance their clinical applicability. RELEVANCE STATEMENT: Deuterium metabolic imaging, a promising non-invasive technique, is systematically discussed in this review for its current progression, limitations, and future directions in studying in vivo energetic metabolism, displaying a relevant clinical potential. KEY POINTS: • Deuterium metabolic imaging (DMI) shows promise for studying in vivo energetic metabolism. • This review explores DMI's current state, limits, and future research directions comprehensively. • The clinical translation of DMI is mainly impeded by limitations in spatial resolution.

Recent advances in microfluidics-based bioNMR analysis.

Nuclear magnetic resonance (NMR) has been used in a variety of fields due to its powerful analytical capability. To facilitate biochemical NMR (bioNMR) analysis for samples with a limited mass, a number of integrated systems have been developed by coupling microfluidics and NMR. However, there are few review papers that summarize the recent advances in the development of microfluidics-based NMR (µNMR) systems. Herein, we review the advancements in µNMR systems built on high-field commercial instruments and low-field compact platforms. Specifically, µNMR platforms with three types of typical microcoils settled in the high-field NMR instruments will be discussed, followed by summarizing compact NMR systems and their applications in biomedical point-of-care testing. Finally, a conclusion and future prospects in the field of µNMR were given.

Functional MRI of murine olfactory bulbs at 15.2T reveals characteristic activation patters when stimulated by different odors.

Thanks to its increased sensitivity, single-shot ultrahigh field functional MRI (UHF fMRI) could lead to valuable insight about subtle brain functions such as olfaction. However, UHF fMRI experiments targeting small organs next to air voids, such as the olfactory bulb, are severely affected by field inhomogeneity problems. Spatiotemporal Encoding (SPEN) is an emerging single-shot MRI technique that could provide a route for bypassing these complications. This is here explored with single-shot fMRI studies on the olfactory bulbs of male and female mice performed at 15.2T. SPEN images collected on these organs at a 108 µm in-plane resolution yielded remarkably large and well-defined responses to olfactory cues. Under suitable T2* weightings these activation-driven changes exceeded 5% of the overall signal intensity, becoming clearly visible in the images without statistical treatment. The nature of the SPEN signal intensity changes in such experiments was unambiguously linked to olfaction, via single-nostril experiments. These experiments highlighted specific activation regions in the external plexiform region and in glomeruli in the lateral part of the bulb, when stimulated by aversive or appetitive odors, respectively. These strong signal activations were non-linear with concentration, and shed light on how chemosensory signals reaching the olfactory epithelium react in response to different cues. Second-level analyses highlighted clear differences among the appetitive, aversive and neutral odor maps; no such differences were evident upon comparing male against female olfactory activation regions.

An easy-built Halbach magnet for LF-NMR with high homogeneity using optimized target-field passive shimming method.

The aim of this work is to develop a Halbach magnet that possesses characteristics such as easy-built, low cost and high homogeneity for use in a portable low-field NMR (LF-NMR) system. Considering portability, a 4-ring Halbach magnet was designed through simulation and mechanical modelling, which was successfully constructed in a general laboratory setting. The obtained field strength (B0) was 0.169 T, with an initial homogeneity of 8204 ppm within a sphere with a diameter of 20 mm. To enhance robustness, efficiency and effectiveness of shimming, an optimized target-field passive shimming method was proposed. Subsequently, the homemade spectrometer was used to run NMR experiments on the Halbach magnet. The 1H NMR linewidths of water samples became significantly narrower after passive shimming, e.g., the linewidth of a sample with a diameter of 3 mm and a length of 3 mm reduced from 452.3 Hz (62.5 ppm) to 12.9 Hz (1.8 ppm), which was much less than 102 Hz. The NMR results demonstrate that the proposed passive shimming method can achieve high homogeneity, and the developed Halbach magnet is capable of satisfying numerous LF-NMR applications.

Correction of Arterial-Phase Motion Artifacts in Gadoxetic Acid-Enhanced Liver MRI Using an Innovative Unsupervised Network.

This study aims to propose and evaluate DR-CycleGAN, a disentangled unsupervised network by introducing a novel content-consistency loss, for removing arterial-phase motion artifacts in gadoxetic acid-enhanced liver MRI examinations. From June 2020 to July 2021, gadoxetic acid-enhanced liver MRI data were retrospectively collected in this center to establish training and testing datasets. Motion artifacts were semi-quantitatively assessed using a five-point Likert scale (1 = no artifact, 2 = mild, 3 = moderate, 4 = severe, and 5 = non-diagnostic) and quantitatively evaluated using the structural similarity index (SSIM) and peak signal-to-noise ratio (PSNR). The datasets comprised a training dataset (308 examinations, including 58 examinations with artifact grade = 1 and 250 examinations with artifact grade ≥ 2), a paired test dataset (320 examinations, including 160 examinations with artifact grade = 1 and paired 160 examinations with simulated motion artifacts of grade ≥ 2), and an unpaired test dataset (474 examinations with artifact grade ranging from 1 to 5). The performance of DR-CycleGAN was evaluated and compared with a state-of-the-art network, Cycle-MedGAN V2.0. As a result, in the paired test dataset, DR-CycleGAN demonstrated significantly higher SSIM and PSNR values and lower motion artifact grades compared to Cycle-MedGAN V2.0 (0.89 ± 0.07 vs. 0.84 ± 0.09, 32.88 ± 2.11 vs. 30.81 ± 2.64, and 2.7 ± 0.7 vs. 3.0 ± 0.9, respectively; p < 0.001 each). In the unpaired test dataset, DR-CycleGAN also exhibited a superior motion artifact correction performance, resulting in a significant decrease in motion artifact grades from 2.9 ± 1.3 to 2.0 ± 0.6 compared to Cycle-MedGAN V2.0 (to 2.4 ± 0.9, p < 0.001). In conclusion, DR-CycleGAN effectively reduces motion artifacts in the arterial phase images of gadoxetic acid-enhanced liver MRI examinations, offering the potential to enhance image quality.

Mutually communicated model based on multi-parametric MRI for automated segmentation and classification of prostate cancer.

BACKGROUND: Multiparametric magnetic resonance imaging (mp-MRI) is introduced and established as a noninvasive alternative for prostate cancer (PCa) detection and characterization. PURPOSE: To develop and evaluate a mutually communicated deep learning segmentation and classification network (MC-DSCN) based on mp-MRI for prostate segmentation and PCa diagnosis. METHODS: The proposed MC-DSCN can transfer mutual information between segmentation and classification components and facilitate each other in a bootstrapping way. For classification task, the MC-DSCN can transfer the masks produced by the coarse segmentation component to the classification component to exclude irrelevant regions and facilitate classification. For segmentation task, this model can transfer the high-quality localization information learned by the classification component to the fine segmentation component to mitigate the impact of inaccurate localization on segmentation results. Consecutive MRI exams of patients were retrospectively collected from two medical centers (referred to as center A and B). Two experienced radiologists segmented the prostate regions, and the ground truth of the classification refers to the prostate biopsy results. MC-DSCN was designed, trained, and validated using different combinations of distinct MRI sequences as input (e.g., T2-weighted and apparent diffusion coefficient) and the effect of different architectures on the network's performance was tested and discussed. Data from center A were used for training, validation, and internal testing, while another center's data were used for external testing. The statistical analysis is performed to evaluate the performance of the MC-DSCN. The DeLong test and paired t-test were used to assess the performance of classification and segmentation, respectively. RESULTS: In total, 134 patients were included. The proposed MC-DSCN outperforms the networks that were designed solely for segmentation or classification. Regarding the segmentation task, the classification localization information helped to improve the IOU in center A: from 84.5% to 87.8% (p < 0.01) and in center B: from 83.8% to 87.1% (p < 0.01), while the area under curve (AUC) of PCa classification was improved in center A: from 0.946 to 0.991 (p < 0.02) and in center B: from 0.926 to 0.955 (p < 0.01) as a result of the additional information provided by the prostate segmentation. CONCLUSION: The proposed architecture could effectively transfer mutual information between segmentation and classification components and facilitate each other in a bootstrapping way, thus outperforming the networks designed to perform only one task.

Simultaneous multi-banding and multi-echo phase encoding for the accelerated acquisition of high-resolution volumetric diffusivity maps by spatiotemporally encoded MRI.

PURPOSE: Spatiotemporal Encoding (SPEN) is an ultrafast imaging technique where the low-bandwidth axis is rasterized in a joint spatial/k-domain. SPEN benefits from increased robustness to field inhomogeneities, folding-free reconstruction of subsampled data, and an ability to combine multiple interleaved or signal averaged scans -yet its relatively high SAR complicates volumetric uses. Here we show how this can be alleviated by merging simultaneous multi-band excitation, with intra-slab multi-echo (ME) phase encoding, for the acquisition of high definition volumetric DWI/DTI data. METHODS: A protocol involving phase-cycling of simultaneous multi-banded z-slab excitations in independently ky-interleaved scans, together with ME trains that kz-encoded positions within these slabs, was implemented. A reconstruction incorporating a CAIPIRINHA-like encoding of the multiple bands and exploiting SPEN's ability to deliver self-referenced, per-shot phase maps, then led to high-definition diffusivity acquisitions, with reduced SAR and acquisition times vis-à-vis non-optimized 3D counterparts. RESULTS: The new protocol was used to collect full brain 3 T DTI experiments at a variety of nominal voxel sizes, ranging from 1.95 to 2.54 mm3. In general, the new protocol yielded superior sensitivity and fewer distortions than what could be observed in comparably timed phase-encoded 3D SPEN, multi-slice 2D SPEN, or optimized EPI counterparts. CONCLUSIONS: A robust procedure for acquiring volumetric DWI/DTI data was developed and demonstrated.

Identification of metabolic kinetic patterns in different brain regions using metabolomics methods coupled with various discriminant approaches.

Metabolomics is widely used as a powerful technique for identifying metabolic patterns and functions of organs and biological systems. Normally, there are multiple groups/targets involved in data processed by discriminant analysis. This is more common in cerebral studies, as there are always several brain regions involved in neuronal studies or brain metabolic dysfunctions. Furthermore, neuronal activity is highly correlated with cerebral energy metabolism, such as oxidation of glucose, especially for glutamatergic (excitatory) and GABAergic (inhibitory) neuronal activities. Thus, regional cerebral energy metabolism recognition is essential for understanding brain functions. In the current study, ten different brain regions were considered for discrimination analysis. The metabolic kinetics were investigated with 13C enrichments in metabolic products of glucose and measured using the nuclear magnetic spectroscopic method. Multiple discriminative methods were used to construct classification models in order to screen out the best method. After comparing all the applied discriminatory analysis methods, the boost-decision tree method was found to be the best method for classification and every cerebral region exhibited its own metabolic pattern. Finally, the differences in metabolic kinetics among these brain regions were analyzed. We, therefore, concluded that the current technology could also be utilized in other multi-class metabolomics studies and special metabolic kinetic patterns could provide useful information for brain function studies.