基于非均匀螺旋线数据和布雷格曼迭代的快速磁共振成像方法

数据采集时间长是制约磁共振成像技术发展的重要瓶颈.为了解决这一问题, 本文基于压缩感知成像理论, 提出了一种结合非均匀螺旋线磁共振数据采集序列和布雷格曼迭代重建的快速磁共振成像方法, 通过欠采样缩短数据采集时间.欠采样引起混迭伪影则通过非均匀螺旋线欠采样特性和布雷格曼迭代重建去除.水模磁共振成像实验和在体磁共振成像实验结果表明: 欠采样情况下, 所提出的方法能有效去除欠采样导致的混迭伪影, 获得的图像结构信息完整的成像结果, 在缩短采样时间的同时, 具有较高的准确度. 关键词： 磁共振成像 非均匀螺旋线 全变分 布雷格曼迭代     

一种以超分辨率理论为基础的磁共振眼球成像方法

磁共振成像（MRI）是一种无电离辐射的非介入性的眼内肿瘤检测方法,但分辨率和运动伪影是成像过程中不易克服的困难．以往的扫描方法或是不可避免的引入运动伪影,或是需要受试者做精确的配合,增加了成像的难度,给受试者带来不舒适的体验．本文提出了一种以超分辨率理论为基础的新的磁共振眼球成像方法,使用一种特制的眼球线圈,对眼部区域扫描一系列动态的图像,使得不同方向上的采集分辨率互补．最后经过预处理、配准、超分辨率重建等操作,得到高质量的磁共振眼球图像．实验结果表明,这种方法可以在不需要受试者做额外配合工作的情况下,得到更加清晰的磁共振眼球图像．     

Micro CT中混叠伪影的消除

Micro CT系统是近几年发展起来的具有极高空间分辨率和探测灵敏度的一种成像技术,专门用于对小动物进行成像,并被广泛应用于药物开发,肿瘤检测.Micro CT的这种高分辨率使存在于CT中的混叠伪影在Micro CT中更为突出,成为影响图像质量,降低图像分辨率的主要因素.本文模拟分析了伪影产生的机理,并提出了一种直接对重构的切片图进行伪影分析及消除的算法,能快速有效的消除Micro CT切片图像里的混叠伪影,伪影消除率达70%以上,并且使图像的对比度提高. 关键词： Micro CT 混叠伪影 自适应滤波     

基于深度递归级联卷积神经网络的并行磁共振成像方法

快速磁共振成像是磁共振研究领域重要的课题之一.随着大数据和深度学习的兴起,神经网络成为快速磁共振技术的重要方法.然而网络性能表现和网络参数量之间较难取得平衡,且对于多通道数据重建的并行成像问题,相关研究较少.本文构建了一种深度递归级联卷积神经网络结构,用于处理并行成像问题.这种网络结构在减少网络参数量的同时,能够尽可能地提高网络的表达能力,提高网络重建的精确度.实验结果表明,相较于传统并行成像方法,通过训练好的神经网络对欠采样磁共振数据进行重建,可以得到更准确的重建结果,且重建时间大大缩短.     

基于差分曲率的各项异性扩散小波图像降噪算法研究

在$\gamma $放射性内污染修正均匀冗余阵列编码成像中,为了进一步减少使用最大似然估计(Maximum Likelihood Expectation Maximisation,MLEM)算法进行编码图像重建的伪影,并且提高重建的效率,本文提出了基于差分曲率的各项异性扩散小波图像降噪(Anisotropic Diffusion Wavelet Image Denoising based on Differential Curvature, 简称ADWIDDC)算法, 结合MLEM算法得到了改进的应用于编码成像的MLEM-ADWIDDC算法。该算法首先采用MLEM算法对探测器得到的投影数据进行图像重建,再利用互补成像方法消减近场伪影,最后采用ADWIDDC算法进一步降低图像伪影。模拟结果表明,对于131I源呈圆环分布的编码图像,以重建算法运行时间为195 s为截止条件,MLEM-ADWIDDC算法能够更好地去除伪影,重建的$\gamma $放射源图像纹理也更加清晰;对于131I源呈“CDUT”分布的编码图像,以重建图像信噪比达到5.10 dB为截止条件,本文的MLEM-ADWIDDC算法运行时间为98.36 s,比仅采用MLEM算法消耗的时间缩短了49%。     

CS-MRI 中评价随机欠采样矩阵的新方法

在压缩感知-磁共振成像(CS-MRI)中,随机欠采样矩阵与重建图像质量密切相关．而选取随机欠采样矩阵一般是通过计算点扩散函数(PSF),以可能产生的伪影的最大值为评价参数,评估欠采样对图像重建的影响,然而最大值只反应了伪影的最坏情况．该文引入了两种新的统计学评价参数平均值(MV)和标准差(SD),其中平均值评估了伪影的平均大小,标准差可以反映伪影的波动情况．该文分别使用这3种参数对小鼠和人体脑部MRI数据以不同的采样比率进行CS图像重建,实验结果表明,当采样比率不低于4倍稀疏度时,使用平均值获得了质量更优的重建图像．因此,通过稀疏度先验知识指导合理选取采样比率,并以平均值为评价参数选取随机欠采样矩阵,能够获得更优的CS-MRI重建图像．
     

压缩感知同步扫描重建及其采样方案的研究

压缩感知(compressed sensing,CS)-磁共振成像(magnetic resonance imaging,MRI)技术使用随机欠采样的k空间数据来重建图像,大大提高了成像速度.但典型的CS重建很费时,这也是CS-MRI临床应用的主要障碍之一.针对这一问题,该文提出了在扫描时同步进行CS图像重建的方案.在同步重建的过程中,可以实时显示重建图像的结果,用户可以根据图像质量来决定何时终止扫描,这样可以在节约扫描和重建时间的同时,更好地控制图像质量.由于预先无法确定最终的采样率,因此传统的变密度随机采样方法并不完全适用.该文设计了适用于同步重建过程的采样模式生成方案,同时提出了分段采样方法,把采样过程分为两个阶段,不同阶段使用不同的概率密度函数(probability density function,PDF)确定待采样的相位编码行.模拟实验的结果表明,与使用单一密度函数的采样方案相比,分段采样方案能够在整个同步扫描重建过程中始终获得更好的图像.     

基于支撑先验与深度图像先验的无预训练磁共振图像重建方法

基于深度学习的磁共振成像(magnetic resonance imaging, MRI)方法需要大规模、高质量的病患数据样本集进行预训练.然而,由于病患隐私及设备等因素限制,获取大规模、高质量的磁共振数据集在实际临床应用中面临挑战.本文提出一种新的基于深度学习的欠采样磁共振图像重建方法,该方法无需预训练、不依赖训练数据集,而是充分利用待重建的目标MR图像的结构先验和支撑先验,并将其引入深度图像先验(deep image prior, DIP)框架,从而削减对训练数据集的依赖,提升学习效率.基于参考图像与目标图像的相似性,采用高分辨率参考图像作为深度网络输入,将结构先验信息引入网络;将参考图像在小波域中幅值大的系数索引集作为目标图像的已知支撑集,构造正则化约束项,将网络训练转化为网络参数的最优化求解过程.实验结果表明,本文方法可由欠采样k空间数据重建得到更精确的磁共振图像,且在保留组织特征、细节纹理方面具有明显优势.     

光学复用大视场成像研究

为了简化光学复用成像系统复杂度和降低对编码方式的要求,基于反射镜旋转完成通道编码,构建了结合分束镜和光学成像系统实现双通道混叠成像的实验装置.为实现混叠图像的有效解混叠,提出了利用特征点检测与匹配方法实现多幅混叠图像之间平移量的自动识别,进而构建出具有亚像素精度的双通道光学复用成像系统的系统复用矩阵,并采用梯度投影稀疏重建算法完成基于多幅图像的双通道图像重建,最终实验实现了光学成像系统视场角的2倍放大.此外,对所提方法在其他场景的适应性也进行了实验验证,表明了其可在不改变焦平面探测器的情况下实现光学成像系统视场角的有效放大,大大节约了大视场光学成像系统的研制成本.     

结构光照明显微中的超分辨图像重建研究

近年来,随着各种新型荧光探针的出现和成像方法的改进,远场光学成像的分辨率已经突破了衍射极限的限制。基于结构光照明的荧光显微技术凭借成像速度快、光毒性弱等优点,已成为目前主流的超分辨成像技术之一。实现结构光照明超分辨显微成像的关键在于照明光场的精准调控和后期的超分辨图像重建算法,否则将会在重建的超分辨图像中产生不可预估的伪影,混淆对观测结构真实形态的判断。详细对比了几种典型的结构光照明显微超分辨重建算法,证明基于图像重组变换的结构光照明超分辨图像重建算法可以有效解决极低结构光场调制度下的超分辨图像重建问题,降低结构光照明显微中的激发光功率。     

ANTHEM: anatomically tailored hexagonal MRI

PURPOSE: This study aimed to investigate the use of anatomically tailored hexagonal sampling for scan-time and error reduction in MRI. MATERIALS AND METHODS: Anatomically tailored hexagonal MRI (ANTHEM), a method that combines hexagonal sampling with specific symmetry in anatomical geometry, is proposed. By using hexagonal sampling, aliasing artifacts are moved to regions where, due to the nature of the anatomy, aliasing is inconsequential. This can be used to either reduce scan time while maintaining spatial resolution or reduce residual errors in speedup techniques like UNFOLD and k-t BLAST/SENSE, which undersample k-space and unwrap fold-over artifacts during reconstruction. Computer simulations as well as phantom and volunteer studies were used to validate the theory. A simplified reconstruction algorithm for hexagonally sampled and subsampled k-space data was also used. RESULTS: A reduction in sampling density of 13.4% and 25% in each hexagonally sampled dimension was achieved for spherical and conical geometries without aliasing or reduction in spatial resolution. Optimal subsampling schemes that can be utilized by UNFOLD and k-t BLAST/SENSE were derived using hexagonal subsampling, which resulted in maximal, isotropic dispersal of the aliases. In combination with UNFOLD, ANTHEM was shown to move residual aliasing artifacts to the corners of the field of view, yielding reduced artifacts in CINE reconstructions. CONCLUSIONS: ANTHEM was successful in reducing acquisition time in conventional MRI and in reducing errors in UNFOLD imaging.     

Solving the problem of concomitant gradients in ultra-low-field MRI

In ultra-low-field magnetic resonance imaging (ULF MRI), spin precession is detected typically in magnetic fields of the order of 10-100 μT. As in conventional high-field MRI, the spatial origin of the signals can be encoded by superposing gradient fields on a homogeneous main field. However, because the main field is weak, gradient field amplitudes become comparable to it. In this case, the concomitant gradients forced by Maxwell's equations cause the assumption of linearly varying field gradients to fail. Thus, image reconstruction with Fourier transformation would produce severe image artifacts. We propose a direct linear inversion (DLI) method to reconstruct images without limiting assumptions about the gradient fields. We compare the quality of the images obtained using the proposed reconstruction method and the Fourier reconstruction. With simulations, we show how the reconstruction errors of the methods depend on the strengths of the concomitant gradients. The proposed approach produces nearly distortion-free images even when the main field reaches zero.     

Deep neural network inspired by iterative shrinkage-thresholding algorithm with data consistency (NISTAD) for fast Undersampled MRI reconstruction

With the aim of developing a fast algorithm for high-quality MRI reconstruction from undersampled k-space data, we propose a novel deep neural Network, which is inspired by Iterative Shrinkage Thresholding Algorithm with Data consistency (NISTAD). NISTAD consists of three consecutive blocks: an encoding block, which models the flow graph of ISTA, a classical iteration-based compressed sensing magnetic resonance imaging (CS-MRI) method; a decoding block, which recovers the image from sparse representation; a data consistency block, which adaptively enforces consistency with the acquired raw data according to learned noise level. The ISTA method is thereby mapped to an end-to-end deep neural network, which greatly reduces the reconstruction time and simplifies the tuning of hyper-parameters, compared to conventional model-based CS-MRI methods. On the other hand, compared to general deep learning-based MRI reconstruction methods, the proposed method uses a simpler network architecture with fewer parameters. NISTAD has been validated on retrospectively undersampled diencephalon standard challenge data using different acceleration factors, and compared with DAGAN and Cascade CNN, two state-of-the-art deep neural network-based methods which outperformed many other state-of-the-art model-based and deep learning-based methods. Experimental results demonstrated that NISTAD reconstruction achieved comparable image quality with DAGAN and Cascade CNN reconstruction in terms of both PSNR and SSIM metrics, and subjective assessment, though with a simpler network structure.     

Transfer learning in deep neural network based under-sampled MR image reconstruction

In Magnetic Resonance Imaging (MRI), the success of deep learning-based under-sampled MR image reconstruction depends on: (i) size of the training dataset, (ii) generalization capabilities of the trained neural network. Whenever there is a mismatch between the training and testing data, there is a need to retrain the neural network from scratch with thousands of MR images obtained using the same protocol. This may not be possible in MRI as it is costly and time consuming to acquire data. In this research, a transfer learning approach i.e. end-to-end fine tuning is proposed for U-Net to address the data scarcity and generalization problems of deep learning-based MR image reconstruction. First the generalization capabilities of a pre-trained U-Net (initially trained on the human brain images of 1.5 T scanner) are assessed for: (a) MR images acquired from MRI scanners of different magnetic field strengths, (b) MR images of different anatomies and (c) MR images under-sampled by different acceleration factors. Later, end-to-end fine tuning of the pre-trained U-Net is proposed for the reconstruction of the above-mentioned MR images (i.e. (a), (b) and (c)). The results show successful reconstructions obtained from the proposed method as reflected by the Structural SIMilarity index, Root Mean Square Error, Peak Signal-to-Noise Ratio and central line profile of the reconstructed images.     

MRISim:磁共振成像仿真实验软件包

本文基于数值计算模拟技术,开发了模拟磁共振成像（MRI）数据采集与图像重建的仿真软件包——MRISim.虚拟数据采集部分通过对设备硬件、样品（标样或人体部位）进行物理数学建模后,构建原始模拟信号,并填充k空间,然后再基于k空间数据实现磁共振图像重建.该软件可以通过参数的任意调节对影响磁共振图像质量的因素进行可视化分析,包括11种常见伪影的特征和成因分析.我们的研究表明应用该仿真软件可以弥补台式MRI扫描仪价格昂贵、实验时间长等不足之处,实现对相关科技人员的批量化、规模化的实践操作培训.     

Referenceless one-dimensional Nyquist ghost correction in multicoil single-shot spatiotemporally encoded MRI

Single-shot spatiotemporally encoded (SPEN) MRI is a novel fast imaging method capable of retaining the time efficiency of single-shot echo planar imaging (EPI) but with distortion artifacts significantly reduced. Akin to EPI, the phase inconsistencies between mismatched even and odd echoes also result in the so-called Nyquist ghosts. However, the characteristic of the SPEN signals provides the possibility of obtaining ghost-free images directly from even and odd echoes respectively, without acquiring additional reference scans. In this paper, a theoretical analysis of the Nyquist ghosts manifested in single-shot SPEN MRI is presented, a one-dimensional correction scheme is put forward capable of maintaining definition of image features without blurring when the phase inconsistency along SPEN encoding direction is negligible, and a technique is introduced for convenient and robust correction of data from multi-channel receiver coils. The effectiveness of the proposed processing pipeline is validated by a series of experiments conducted on simulation data, in vivo rats and healthy human brains. The robustness of the method is further verified by implementing distortion correction on ghost corrected data.     

Development of fast deep learning quantification for magnetic resonance fingerprinting in vivo

PurposeA deep neural network was developed for magnetic resonance fingerprinting (MRF) quantification. This study aimed at extending previous studies of deep learning MRF to in vivo applications, allowing sub-second computation time for large-scale data.MethodsWe applied the deep learning methodology based on our previously published multi-layer perceptron. The number of layers was four, which was optimized to balance the model capacity and noise robustness. The training sets were obtained from MRF dictionaries with 9000 to 28,000 atoms, depending on the desired T1 and T2 ranges. The simulated MRF undersampling artifact based on the k-space acquisition scheme and noise were both added to the training data to reduce the error in estimates.ResultsThe neural network achieved high fidelity (R2 _ 0.98) as compared to the T1 and T2 values of the ISMRM standardized phantom. In brain MRF experiment, the model trained with simulated artifacts and noise showed less error compared to that without. The in vivo application of our neural network for liver and prostate were also demonstrated. For an MRF slice with 256 _ 256 image resolution, the computation time of our neural network was 0.12 s, compared with the _ 28 s-pre-slice for the conventional dictionary matching method.ConclusionOur neural network achieved fast computation speed for MRF quantification. The model trained with simulated artifacts and noise showed less error and achieved optimal performance for phantom experiment and in vivo normal brain and liver, and prostate cancer patient.     

Super-resolved parallel MRI by spatiotemporal encoding

Recent studies described an “ultrafast” scanning method based on spatiotemporal (SPEN) principles. SPEN demonstrates numerous potential advantages over EPI-based alternatives, at no additional expense in experimental complexity. An important aspect that SPEN still needs to achieve for providing a competitive ultrafast MRI acquisition alternative, entails exploiting parallel imaging algorithms without compromising its proven capabilities. The present work introduces a combination of multi-band frequency-swept pulses simultaneously encoding multiple, partial fields-of-view, together with a new algorithm merging a Super-Resolved SPEN image reconstruction and SENSE multiple-receiving methods. This approach enables one to reduce both the excitation and acquisition times of sub-second SPEN acquisitions by the customary acceleration factor R, without compromises in either the method’s spatial resolution, SAR deposition, or capability to operate in multi-slice mode. The performance of these new single-shot imaging sequences and their ancillary algorithms were explored and corroborated on phantoms and human volunteers at 3 T. The gains of the parallelized approach were particularly evident when dealing with heterogeneous systems subject to major T₂/T₂* effects, as is the case upon single-scan imaging near tissue/air interfaces.     

Optimizing Image Reconstruction in SENSE Using GPU

Parallel magnetic resonance imaging (MRI) (pMRI) uses multiple receiver coils to reduce the MRI scan time. To accelerate the data acquisition process in MRI, less amount of data is acquired from the scanner which leads to artifacts in the reconstructed images. SENSitivity Encoding (SENSE) is a reconstruction algorithm in pMRI to remove aliasing artifacts from the undersampled multi coil data and recovers fully sampled images. The main limitation of SENSE is computing inverse of the encoding matrix. This work proposes the inversion of encoding matrix using Jacobi singular value decomposition (SVD) algorithm for image reconstruction on GPUs to accelerate the reconstruction process. The performance of Jacobi SVD is compared with Gauss–Jordan algorithm. The simulations are performed on two datasets (brain and cardiac) with acceleration factors 2, 4, 6 and 8. The results show that the graphics processing unit (GPU) provides a speed up to 21.6 times as compared to CPU reconstruction. Jacobi SVD algorithm performs better in terms of acceleration in reconstructions on GPUs as compared to Gauss–Jordan method. The proposed algorithm is suitable for any number of coils and acceleration factors for SENSE reconstruction on real time processing systems.