3D MRI-based tumor delineation of ocular melanoma and its comparison with conventional techniques

The aim of this study is to (1) compare the delineation of the tumor volume for ocular melanoma on high-resolution three-dimensional (3D) T2-weighted fast spin echo magnetic resonance imaging (MRI) images with conventional techniques of A- and B-scan ultrasound, transcleral illumination, and placement of tantalum markers around tumor base and (2) to evaluate whether the surgically placed marker ring tumor delineation can be replaced by 3D MRI based tumor delineation. High-resolution 3D T2-weighted fast spin echo (3D FSE) MRI scans were obtained for 60 consecutive ocular melanoma patients using a 1.5 T MRI (GE Medical Systems, Milwaukee, WI), in a standard head coil. These patients were subsequently treated with proton beam therapy at the UC Davis Cyclotron, Davis, CA. The tumor was delineated by placement of tantalum rings (radio-opaque markers) around the tumor periphery as defined by pupillary transillumination during surgery. A point light source, placed against the sclera, was also used to confirm ring agreement with indirect ophthalmoscopy. When necessary, intraoperative ultrasound was also performed. The patients were planned using EYEPLAN software and the tumor volumes were obtained. For analysis, the tumors were divided into four categories based on tumor height and basal diameter. In order to assess the impact of high-resolution 3D T2 FSE MRI, the tumor volumes were outlined on the MRI scans by two independent observers and the tumor volumes calculated for each patient. Six (10%) of 60 patients had tumors, which were not visible on 3D MRI images. These six patients had tumors with tumor heights < or = 3 mm. A small intraobserver variation with a mean of (-0.22 +/- 4)% was seen in tumor volumes delineated by 3D T2 FSE MR images. The ratio of tumor volumes measured on MRI to EYEPLAN for the largest to the smallest tumor volumes varied between 0.993 and 1.02 for 54 patients. The tumor volumes measured directly on 3D T2 FSE MRI ranged from 4.03 to 0.075 cm3. with a mean of 0.87 +/- 0.84 cm3. The tumor shapes obtained from 3D T2 FSE MR images were comparable to the tumor shapes obtained using EYEPLAN software. The demonstration of intraocular tumor volumes with the high-resolution 3D fast spin echo T2 weighted MRI is excellent and provides additional information on tumor shape. We found a high degree of accuracy for tumor volumes with direct MRI volumetric measurements in uveal melanoma patients. In some patients with extra large tumors, the tumor base and shape was modified, because of the additional information obtained from 3D T2 FSE MR images.     

基于对抗训练的U-Net神经网络在稀疏投影CT图像增强的应用

目的:针对稀疏投影的CT重建图像附带噪声和伪影的特性,使用神经网络模型对稀疏投影得到的低质量CT重建 图像进行图像增强。方法:在残差编码-解码卷积神经网络基础上提出一种基于对抗训练的U-Net神经网络模型,并使用 公开数据集TCGA-CESC癌症CT影像进行模型训练和测试。评价模型处理效果的指标包括峰值信噪比（PSNR）、结构相 似性（SSIM）和均方根误差（RMSE）。结果:在对180 次探测的CT重建图像的测试中,模型处理后的图像相比未处理图 像,PSNR、SSIM和RMSE指标平均值分别提升15.10%、37.89%和38.20%。在PSNR和SSIM指标平均值意义下,模型处 理后的图像优于1 800次探测的未处理CT重建图像。结论:本研究提出的神经网络模型能够减少伪影和噪点,对稀疏投 影CT图像增强有一定效果。     

用于乳腺超声透射断层成像的飞行时间精确计算方法

目的:针对乳腺超声透射断层成像（UTCT），提出一种精确计算超声波飞行时间（TOF）的方法，实现定量乳腺声速图像的高质量重建。方法:利用k-wave平台设置超声透射断层扫描环境，获取两种不同复杂度的数字乳腺体模的扫描数据，分别使用提出的极值点法及常用的阈值法、过零点法和峰值法计算出TOF，并借助滤波反投影算法重建声速图像。结果:峰值法计算出的TOF难以重建出完整的声速图像，极值点法重建出的声速图像在均方根误差（RMSE）和结构相似性指标（SSIM）上均优于阈值法和过零点法，相比于阈值法和过零点法，RMSE平均减小了8.025%、1.643%， SSIM平均增加了1.708%、1.68%。结论:本文提出的极值点法能够精确计算出TOF，有助于UTCT进行高质量的声速图像重建。     

基于神经网络学习的锥形束CT图像超分辨率重建算法

针对锥形束CT（CBCT）图像质量较差的问题,提出一种基于卷积神经网络的超分辨率重建（SRCNN）方法,旨在提高CBCT图像的分辨率。本研究分别对头颈、盆腔、胸部的CBCT图像进行研究,先使用非局部均值（NLM）方法对图像进行降噪处理,再分别使用双三次插值重建（BIC）方法和SRCNN重建方法进行超分辨率重建。结果表明,BIC方法和SRCNN重建方法均能提高CBCT图像的分辨率,SRCNN重建方法较BIC方法有更高的峰值信噪比,而在结构相似度和特征相似度上,BIC方法和SRCNN重建方法的差别不大。从图像峰值信噪比及特征相似度上看,此方法对盆腔部CBCT图像处理效果更为显著,对头颈部及胸部处理效果相近。     

Rapid reconstruction of highly undersampled,non‐Cartesian real‐time cine k‐space data using a perceptual complex neural network (PCNN)

Highly accelerated real‐time cine MRI using compressed sensing (CS) is a promising approach to achieve high spatio‐temporal resolution and clinically acceptable image quality in patients with arrhythmia and/or dyspnea. However, its lengthy image reconstruction time may hinder its clinical translation. The purpose of this study was to develop a neural network for reconstruction of non‐Cartesian real‐time cine MRI k‐space data faster (<1 min per slice with 80 frames) than graphics processing unit (GPU)‐accelerated CS reconstruction, without significant loss in image quality or accuracy in left ventricular (LV) functional parameters. We introduce a perceptual complex neural network (PCNN) that trains on complex‐valued MRI signal and incorporates a perceptual loss term to suppress incoherent image details. This PCNN was trained and tested with multi‐slice, multi‐phase, cine images from 40 patients (20 for training, 20 for testing), where the zero‐filled images were used as input and the corresponding CS reconstructed images were used as practical ground truth. The resulting images were compared using quantitative metrics (structural similarity index (SSIM) and normalized root mean square error (NRMSE)) and visual scores (conspicuity, temporal fidelity, artifacts, and noise scores), individually graded on a five‐point scale (1, worst; 3, acceptable; 5, best), and LV ejection fraction (LVEF). The mean processing time per slice with 80 frames for PCNN was 23.7 ± 1.9 s for pre‐processing (Step 1, same as CS) and 0.822 ± 0.004 s for dealiasing (Step 2, 166 times faster than CS). Our PCNN produced higher data fidelity metrics (SSIM = 0.88 ± 0.02, NRMSE = 0.014 ± 0.004) compared with CS. While all the visual scores were significantly different (P < 0.05), the median scores were all 4.0 or higher for both CS and PCNN. LVEFs measured from CS and PCNN were strongly correlated (R² = 0.92) and in good agreement (mean difference = ?1.4% [2.3% of mean]; limit of agreement = 10.6% [17.6% of mean]). The proposed PCNN is capable of rapid reconstruction (25 s per slice with 80 frames) of non‐Cartesian real‐time cine MRI k‐space data, without significant loss in image quality or accuracy in LV functional parameters.     

Interactively variable isotropic resolution in computed tomography

An individual balancing between spatial resolution and image noise is necessary to fulfil the diagnostic requirements in medical CT imaging. In order to change influencing parameters, such as reconstruction kernel or effective slice thickness, additional raw-data-dependent image reconstructions have to be performed. Therefore, the noise versus resolution trade-off is time consuming and not interactively applicable. Furthermore, isotropic resolution, expressed by an equivalent point spread function (PSF) in every spatial direction, is important for the undistorted visualization and quantitative evaluation of small structures independent of the viewing plane. Theoretically, isotropic resolution can be obtained by matching the in-plane and through-plane resolution with the aforementioned parameters. Practically, however, the user is not assisted in doing so by current reconstruction systems and therefore isotropic resolution is not commonly achieved, in particular not at the desired resolution level. In this paper, an integrated approach is presented for equalizing the in-plane and through-plane spatial resolution by image filtering. The required filter kernels are calculated from previously measured PSFs in x/y- and z-direction. The concepts derived are combined with a variable resolution filtering technique. Both approaches are independent of CT raw data and operate only on reconstructed images which allows for their application in real time. Thereby, the aim of interactively variable, isotropic resolution is achieved. Results were evaluated quantitatively by measuring PSFs and image noise, and qualitatively by comparing the images to direct reconstructions regarded as the gold standard. Filtered images matched direct reconstructions with arbitrary reconstruction kernels with standard deviations in difference images of typically between 1 and 17 HU. Isotropic resolution was achieved within 5% of the selected resolution level. Processing times of 20-100 ms per frame allow for interactive use.     

Inline dual-echo T2 quantification in brain using a fast mapping reconstruction technique

T2 mapping from 2D proton density and T2–weighted images (PD-T2) using Bloch equation simulations can be time consuming and introduces a latency between image acquisition and T2 map production. A fast T2 mapping reconstruction method is investigated and compared with a previous modeling approach to reduce computation time and allow inline T2 maps on the MRI console. Brain PD-T2 images from five multiple sclerosis patients were used to compare T2 map reconstruction times between the new subtraction method and the Euclidean norm minimization technique. Bloch equation simulations were used to create the lookup table for decay curve matching in both cases. Agreement of the two techniques used Bland–Altman analysis for investigating individual subsets of data and all image points in the five volumes (meta-analysis). The subtraction method resulted in an average reduction of computation time for single slices from 134 s (minimization method) to 0.44 s. Comparing T2 values between the subtraction and minimization methods resulted in a confidence interval ranging from −0.06 to 0.06 ms (95% of values were within ± 0.06 ms between the techniques). Using identical reconstruction code based on the subtraction method, inline T2 maps were produced from PD-T2 images directly on the scanner console. The excellent agreement between the two methods permits the subtraction technique to be interchanged with the previous method, reducing computation time and allowing inline T2 map reconstruction based on Bloch simulations directly on the scanner.     

A three-dimensional reconstruction algorithm for an inverse-geometry volumetric CT system

An inverse-geometry volumetric computed tomography (IGCT) system has been proposed capable of rapidly acquiring sufficient data to reconstruct a thick volume in one circular scan. The system uses a large-area scanned source opposite a smaller detector. The source and detector have the same extent in the axial, or slice, direction, thus providing sufficient volumetric sampling and avoiding cone-beam artifacts. This paper describes a reconstruction algorithm for the IGCT system. The algorithm first rebins the acquired data into two-dimensional (2D) parallel-ray projections at multiple tilt and azimuthal angles, followed by a 3D filtered backprojection. The rebinning step is performed by gridding the data onto a Cartesian grid in a 4D projection space. We present a new method for correcting the gridding error caused by the finite and asymmetric sampling in the neighborhood of each output grid point in the projection space. The reconstruction algorithm was implemented and tested on simulated IGCT data. Results show that the gridding correction reduces the gridding errors to below one Hounsfield unit. With this correction, the reconstruction algorithm does not introduce significant artifacts or blurring when compared to images reconstructed from simulated 2D parallel-ray projections. We also present an investigation of the noise behavior of the method which verifies that the proposed reconstruction algorithm utilizes cross-plane rays as efficiently as in-plane rays and can provide noise comparable to an in-plane parallel-ray geometry for the same number of photons. Simulations of a resolution test pattern and the modulation transfer function demonstrate that the IGCT system, using the proposed algorithm, is capable of 0.4 mm isotropic resolution. The successful implementation of the reconstruction algorithm is an important step in establishing feasibility of the IGCT system.     

Quantitative image quality evaluation of MR images using perceptual difference models

The authors are using a perceptual difference model (Case-PDM) to quantitatively evaluate image quality of the thousands of test images which can be created when optimizing fast magnetic resonance (MR) imaging strategies and reconstruction techniques. In this validation study, they compared human evaluation of MR images from multiple organs and from multiple image reconstruction algorithms to Case-PDM and similar models. The authors found that Case-PDM compared very favorably to human observers in double-stimulus continuous-quality scale and functional measurement theory studies over a large range of image quality. The Case-PDM threshold for nonperceptible differences in a 2-alternative forced choice study varied with the type of image under study, but was approximately 1.1 for diffuse image effects, providing a rule of thumb. Ordering the image quality evaluation models, we found in overall Case-PDM approximately IDM (Sarnoff Corporation) approximately SSIM [Wang et al. IEEE Trans. Image Process. 13, 600-612 (2004)] > mean squared error NR [Wang et al. (2004) (unpublished)] > DCTune (NASA) > IQM (MITRE Corporation). The authors conclude that Case-PDM is very useful in MR image evaluation but that one should probably restrict studies to similar images and similar processing, normally not a limitation in image reconstruction studies.     

基于深度学习网络的医学核磁共振成像超分辨率重构实验

基于深度学习网络的医学核磁共振（MR）图像超分辨重建实验研究,提出并构建一个大规模的高质量用于MR图像超分辨的数据集,涵盖了头颅、膝盖、乳房以及头颈4个部位。通过数据质量筛选和不同低分辨率图像生成方式,在原始图像的高分辨率基础下,以×2、×3、×4的下采样尺度,原始MRI图像形成3种不同尺度下的MR图像数据集,同时给出不同部位超分辨难易程度分析。采用7个在自然图像的超分辨率领域中取得最好效果的深度学习网络,将它们迁移到MR图像中,学习低分辨率MR图像到高低分辨MR图像的映射关系,并对比分析这些深度学习网络在自然图像的超分辨效果。通过实验可以看出,深度学习网络在MR图像超分辨取得了比传统算法更好的效果,部分结果不亚于自然图像;不同部位的超分辨效果差异较大,难以以一个深度学习网络使不同部位均具有更好的超分辨效果。深度学习网络在MR图像超分辨将具有重要的应用价值和理论意义。     

Robust-Deep: A Method for Increasing Brain Imaging Datasets to Improve Deep Learning Models’ Performance and Robustness

A small dataset commonly affects generalization, robustness, and overall performance of deep neural networks (DNNs) in medical imaging research. Since gathering large clinical databases is always difficult, we proposed an analytical method for producing a large realistic/diverse dataset. Clinical brain PET/CT/MR images including full-dose (FD), low-dose (LD) corresponding to only 5 % of events acquired in the FD scan, non-attenuated correction (NAC) and CT-based measured attenuation correction (MAC) PET images, CT images and T1 and T2 MR sequences of 35 patients were included. All images were registered to the Montreal Neurological Institute (MNI) template. Laplacian blending was used to make a natural presentation using information in the frequency domain of images from two separate patients, as well as the blending mask. This classical technique from the computer vision and image processing communities is still widely used and unlike modern DNNs, does not require the availability of training data. A modified ResNet DNN was implemented to evaluate four image-to-image translation tasks, including LD to FD, LD+MR to FD, NAC to MAC, and MRI to CT, with and without using the synthesized images. Quantitative analysis using established metrics, including the peak signal-to-noise ratio (PSNR), structural similarity index metric (SSIM), and joint histogram analysis was performed for quantitative evaluation. The quantitative comparison between the registered small dataset containing 35 patients and the large dataset containing 350 synthesized plus 35 real dataset demonstrated improvement of the RMSE and SSIM by 29% and 8% for LD to FD, 40% and 7% for LD+MRI to FD, 16% and 8% for NAC to MAC, and 24% and 11% for MRI to CT mapping task, respectively. The qualitative/quantitative analysis demonstrated that the proposed model improved the performance of all four DNN models through producing images of higher quality and lower quantitative bias and variance compared to reference images.     

基于频域约束和交叉融合特征网络的磁共振图像超分辨率重建算法

磁共振（MR）图像常用于临床医学诊断，获得高分辨率MR图像有利于进行医学分析。目前主流的基于参考的图像超分辨率重建算法重建的图像，其视觉效果取得了明显的提升，但仍存在明显的伪影问题。针对该问题，提出频域约束和交叉融合特征网络（FCCF）模型，即引入频域损失函数作为约束条件，并构建一种多分辨率特征融合机制，通过交叉融合不同分辨率的图像特征来提高生成图像的质量，使重建结果具有更清晰的细节，没有明显的伪影。在合成和真实的MR图像数据集上分别用PSNR和SSIM指标进行评估，实验结果明显优于现有的超分辨率重建方法。     

Deep complex convolutional network for fast reconstruction of 3D late gadolinium enhancement cardiac MRI

Several deep‐learning models have been proposed to shorten MRI scan time. Prior deep‐learning models that utilize real‐valued kernels have limited capability to learn rich representations of complex MRI data. In this work, we utilize a complex‐valued convolutional network (?Net ) for fast reconstruction of highly under‐sampled MRI data and evaluate its ability to rapidly reconstruct 3D late gadolinium enhancement (LGE) data. ?Net preserves the complex nature and optimal combination of real and imaginary components of MRI data throughout the reconstruction process by utilizing complex‐valued convolution, novel radial batch normalization, and complex activation function layers in a U‐Net architecture. A prospectively under‐sampled 3D LGE cardiac MRI dataset of 219 patients (17 003 images) at acceleration rates R = 3 through R = 5 was used to evaluate ?Net . The dataset was further retrospectively under‐sampled to a maximum of R = 8 to simulate higher acceleration rates. We created three reconstructions of the 3D LGE dataset using (1) ?Net , (2) a compressed‐sensing‐based low‐dimensional‐structure self‐learning and thresholding algorithm (LOST), and (3) a real‐valued U‐Net (realNet) with the same number of parameters as ?Net . LOST‐reconstructed data were considered the reference for training and evaluation of all models. The reconstructed images were quantitatively evaluated using mean‐squared error (MSE) and the structural similarity index measure (SSIM), and subjectively evaluated by three independent readers. Quantitatively, ?Net ‐reconstructed images had significantly improved MSE and SSIM values compared with realNet (MSE, 0.077 versus 0.091; SSIM, 0.876 versus 0.733, respectively; p < 0.01). Subjective quality assessment showed that ?Net ‐reconstructed image quality was similar to that of compressed sensing and significantly better than that of realNet. ?Net reconstruction was also more than 300 times faster than compressed sensing. Retrospective under‐sampled images demonstrate the potential of ?Net at higher acceleration rates. ?Net enables fast reconstruction of highly accelerated 3D MRI with superior performance to real‐valued networks, and achieves faster reconstruction than compressed sensing.     

Positive contrast of SPIO‐labeled cells by off‐resonant reconstruction of 3D radial half‐echo bSSFP

This article describes a new acquisition and reconstruction concept for positive contrast imaging of cells labeled with superparamagnetic iron oxides (SPIOs). Overcoming the limitations of a negative contrast representation as gained with gradient echo and fully balanced steady state (bSSFP), the proposed method delivers a spatially localized contrast with high cellular sensitivity not accomplished by other positive contrast methods. Employing a 3D radial bSSFP pulse sequence with half‐echo sampling, positive cellular contrast is gained by adding artificial global frequency offsets to each half‐echo before image reconstruction. The new contrast regime is highlighted with numerical intravoxel simulations including the point‐spread function for 3D half‐echo acquisitions. Furthermore, the new method is validated on the basis of in vitro cell phantom measurements on a clinical MRI platform, where the measured contrast‐to‐noise ratio (CNR) of the new approach exceeds even the negative contrast of bSSFP. Finally, an in vivo proof of principle study based on a mouse model with a clear depiction of labeled cells within a subcutaneous cell islet containing a cell density as low as 7 cells/mm³ is presented. The resultant isotropic images show robustness to motion and a high CNR, in addition to an enhanced specificity due to the positive contrast of SPIO‐labeled cells. Copyright © 2014 John Wiley & Sons, Ltd.     

Statistical reconstruction for x-ray CT systems with non-continuous detectors

We analyse the performance of statistical reconstruction (SR) methods when applied to non-continuous x-ray detectors. Robustness to projection gaps is required in x-ray CT systems with multiple detector modules or with defective detector pixels. In such situations, the advantage of statistical reconstruction is that it is able to ignore missing or faulty pixels and that it makes optimal use of the remaining line integrals. This potentially obviates the need to fill the sinogram discontinuities by interpolation or any other approximative pre-processing techniques. In this paper, we apply SR to cone beam projections of (i) a hypothetical modular detector micro-CT scanner and of (ii) a system with randomly located defective detector elements. For the modular-detector system, SR produces reconstruction volumes free of noticeable gap-induced artefacts as long as the location of detector gaps and selection of the scanning range provide complete object sampling in the central imaging plane. When applied to randomly located faulty detector elements, SR produces images free of substantial ring artefacts even for cases where defective pixels cover as much as 3% of the detector area.     

Dual-contrast echo planar imaging with keyhole: application to dynamic contrast-enhanced perfusion studies

A new EPI-based method is presented which features optimized sampling of k-space enabling the integrated acquisition of two gradient echo images. The first of these images is predominantly T1 weighted and the second is T*2 weighted. The new method combines echo sharing of sparsely acquired high spatial frequency components with the keyhole technique and half-Fourier image reconstruction. The feasibility of acquiring high spatial and temporal resolution in vivo images for perfusion mapping is demonstrated. In contrast to most current perfusion methods, which acquire the T1- and T*2-weighted images in separate acquisitions, the need for image co-registration here is obviated since both sets of images are EPI-based and are acquired within the same measurement.     

Spatial reconstruction of coronary arteries from angiographic images

A method for 3D reconstruction of coronary arteries from two angiographic projection views is presented (cinefilm 35 mm, 50 images/s). The geometric information defining the two projections is secured by filming a cube bearing 15 steel markers in the approximate location of the heart. For 3D reconstruction, a pair of images of the cube and two simultaneous images of the coronary tree are digitized on a computer assisted system. Two matrices describing the two projections are obtained from the 15 x 2 x 2 image coordinates of the cube markers. Next, the operator draws with a mouse the approximate centre lines of the coronary arteries to be 3D reconstructed. The program then determines the centre lines and the edges of the vessels in the two images accurately. The 3D centre lines are reconstructed and the true local vessel diameters determined. Intravascular volumes are obtained by summation of elementary volumes. The reconstructed arteries are visualized on a colour TV monitor. Measurements on various phantoms have shown that lengths of a few centimetres can be measured with typically 2% error. The accuracy of volume determination was somewhat less due to the inaccuracy of lumen determination. For real coronary arteries the volume error is estimated to be about 15% because of the poorer image quality.     

Evaluation of congenital heart disease by three-dimensional magnetic resonance imaging

This study was undertaken to compare electrocardiographically gated magnetic resonance imaging (MRI) to established imaging modalities in the detection of complex intra- and extracardiac morphologic defects. Twenty-three patients with congenital cardiac abnormalities were imaged by four methods: cardiac catheterization, echocardiography, two-dimensional (2D) transaxial MRI, and three-dimensional (3D) MRI surface reconstruction. Observers with experience in congenital cardiac disease diagnosis (two for echo, one for catheterization, two for 2D MR, and three for 3D MR) evaluated the images in a blinded fashion, and the results were analyzed with receiver operating characteristic (ROC) analysis. Overall, cardiac catheterization had the best diagnostic performance. The diagnostic value of routine 2D cardiac MR images and 3D MR reconstruction images were similar to that of echocardiography. All of the modalities performed poorly in the diagnosis of extracardiac defects and atrial septal defects.     

High resolution 3.0 Tesla MR imaging findings in patients with bilateral Madelung’s deformity

Purpose  Madelung deformity (MD) is a rare, normally painful abnormality of the wrist and forearm which characteristically begins in adolescence. Usually the deformity appears between the age of 8 and 14 years, often progressing from initially mild functional pain to fatigue and loss of strength and finally, reduced mobility. We present the MR-findings in three patients with bilateral MD, using a high-resolution imaging protocol adapted for 3.0 Tesla (3.0 T) examinations. Materials and methods  Wrist images of three patients were acquired at a 3.0 T Scanner (Gyroscan Intera, Philips Medical Systems, Best, The Netherlands), using a dedicated phased array coil. The imaging protocol consisted of coronal T1-weighted Turbo-spin-echo (T1w-TSE) and coronal and sagittal T2-weighted TSE sequences (T2w-TSE). Results  MR-images of these three girls demonstrated severe volar bayonet configuration of the forearms with a dorsal prominence of the ulnar head, also a curved distal radial articular surface with increased ulnar angulation, due to a deceleration of growth in the ulnar portion of the distal epiphysis. The proximal carpal row showed pyramidal configuration. Also visible was a prominent short radiolunate ligament, the so called Vickers ligament, which originates from the ulnar border of the radius, inserts into the volar pole of the lunate and likely contributes to carpal pyramidalization. Furthermore, the images demonstrated an anomalous hypertrophied and elongated volar radiotriquetral ligament which, to our knowledge, has been described elsewhere only in another case. Conclusion  High resolution imaging at 3.0 T permitted a detailed analysis of the complex pathomorphology in patients with MD. Investing the better signal-to-noise ratio at higher field strengths into spatial resolution an excellent image quality could be obtained, depicting the Vickers ligament and the anomalous volar radiotriquetral ligament in this rare disease.     

DEep learning-based rapid Spiral Image REconstruction (DESIRE) for high-resolution spiral first-pass myocardial perfusion imaging

The objective of the current study was to develop and evaluate a DEep learning-based rapid Spiral Image REconstruction (DESIRE) technique for high-resolution spiral first-pass myocardial perfusion imaging with whole-heart coverage, to provide fast and accurate image reconstruction for both single-slice (SS) and simultaneous multislice (SMS) acquisitions. Three-dimensional U-Net–based image enhancement architectures were evaluated for high-resolution spiral perfusion imaging at 3 T. The SS and SMS MB = 2 networks were trained on SS perfusion images from 156 slices from 20 subjects. Structural similarity index (SSIM), peak signal-to-noise ratio (PSNR), and normalized root mean square error (NRMSE) were assessed, and prospective images were blindly graded by two experienced cardiologists (5: excellent; 1: poor). Excellent performance was demonstrated for the proposed technique. For SS, SSIM, PSNR, and NRMSE were 0.977 [0.972, 0.982], 42.113 [40.174, 43.493] dB, and 0.102 [0.080, 0.125], respectively, for the best network. For SMS MB = 2 retrospective data, SSIM, PSNR, and NRMSE were 0.961 [0.950, 0.969], 40.834 [39.619, 42.004] dB, and 0.107 [0.086, 0.133], respectively, for the best network. The image quality scores were 4.5 [4.1, 4.8], 4.5 [4.3, 4.6], 3.5 [3.3, 4], and 3.5 [3.3, 3.8] for SS DESIRE, SS L1-SPIRiT, MB = 2 DESIRE, and MB = 2 SMS-slice-L1-SPIRiT, respectively, showing no statistically significant difference (p = 1 and p = 1 for SS and SMS, respectively) between L1-SPIRiT and the proposed DESIRE technique. The network inference time was ~100 ms per dynamic perfusion series with DESIRE, while the reconstruction time of L1-SPIRiT with GPU acceleration was ~ 30 min. It was concluded that DESIRE enabled fast and high-quality image reconstruction for both SS and SMS MB = 2 whole-heart high-resolution spiral perfusion imaging.