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.     

基于深度反向投影的感知增强超分辨率重建模型

以SRCNN(super-resolution convolutional neural network)模型为代表的超分辨率重建模型通常都有很高的PSNR(peak signal to noise ratio)和SSIM(structural similarity)值,但其在视觉感知上并不令人满意,而以SRGAN为代...     

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．     

新型超高时空分辨率超分辨成像技术

众所周知,白光通过三棱镜后可以分为赤橙黄绿青蓝紫的七色光,每种颜色代表了一段光谱。最近,北京大学席鹏课题组利用量子点的光谱特性,成功实现了新型超高时空分辨率的超分辨成像技术。相关研究成果发表在Nature出版集团的《科学报告》杂志上。超分辨技术(super-resolution)因其能够提供低于衍射极限的分辨率而获得了2014年诺     

基于自适应正则化的超分辨率重建方法

传统正则化超分辨重建得到的图像往往存在过度平滑或伪信息残留的问题,结合超分辨重建模型对重建图像伪信息的产生进行了分析,针对传统方法的不足提出了基于图像区域信息自适应的正则化方法,通过图像的区域信息将图像划分为平滑区与非平滑区域,对不同区域选用不同的先验模型进行约束。同时考虑人眼的视觉感知特性,结合区域信息实现正则化参数的自适应选取。实验结果表明该方法在抑制重建图像伪信息的同时能有效保护细节,效果要优于传统方法与单一的先验模型约束,对于红外与可见光图像重建效果的提升提供了一定的理论参考。     

频谱编码深度成像实验研究

频谱编码显微镜是用衍射光栅和光谱分析装置来获得显微图像.样品上不同的位置被不同的波长照明,通过对反射光光谱进行解码来得到空间信息.搭建了一个基于超连续光源和自制光谱仪的频谱编码显微成像系统,其横向分辨率为1.72±0.13μm(编码线方向)和1.26±0.08μm(垂直于编码线方向),测得不同横向位置处的轴向分辨率有差异.对离体猪肝组织不同部位进行了成像(可见血管、肝窦内皮细胞和肝细胞);对鸡心组织以10μm深度间隔进行成像,测得不同深度处结构信息不一样.结果表明,采用该频谱编码成像的方法能够进行高分辨的深度成像.     

基于宽带立体超透镜的远场超分辨率成像

为突破传统衍射极限实现远场超分辨率成像,提出了一种微波频段宽带立体超透镜用于目标远场超分辨率成像.该透镜可将携带着目标超分辨率信息的凋落波分量转换为传播波分量辐射到远场,进而可在远场接收这些信息并用于超分辨率成像.分别从频域和时域两方面对该透镜的超分辨率特性进行验证.在频域,利用多重信号分类算法对借助于该结构的扩展目标实现了λ/12的远场超分辨率成像,大幅度提升了成像效果.在时域,结合时间反演技术,验证了带宽提升对空间超分辨率聚焦特性带来的明显优势.     

基于水下成像系统模型的水下图像超分辨率重建技术研究

水下成像技术在诸多领域获得了越来越多的应用,然而由于受到成像器件参数、水体特性等成像系统参数的影响,水下图像的分辨率普遍较低、像质较差。基于包括点扩散函数、衍射极限等水下成像系统模型的图像超分辨率重建技术,能够在提高图像分辨率的同时增强图像质量。为了尽可能提高图像分辨率,建立了基于光束传播理论的超分辨率成像模型,并将其应用于水下脉冲激光距离选通成像结果图像的超分辨率重构。重构实验的结果表明,所提出的方法可以有效地提高水下成像的分辨率和质量。     

基于超分辨率重建的图像增强算法研究

 为了提高图像分辨率,从软件的角度出发,对低分辨率图像序列重建高分辨率图像的原理和算法实现开展了研究,提出从多个低分辨率图像序列中获取更高分辨率图像的方法。通过采用基于光流场的金字塔分层结构实现由粗到精的图像配准,获取了亚像素的运动估计。在采用多帧低分辨率图像进行亚像素级配准后,提出采用动态自适应确定正则化参数的方法,构造了简单的正则化代价函数,建立了低分辨率图像与高分辨率图像之间关系的重建模型,仿真实验结果表明该超分辨率重建算法水平和垂直方向上分辨率各增加一倍,与其他算法相比清晰度更高,计算时间不到传统POCS算法的一半。     

数字重建编码成像的迭代算法

用矩阵方法研究数字重建编码成像的迭代方法，为克服问题的不唯一与病态，采用正规化方法，给出正视化迭代解码算法，分析了算法的收敛性，与真解的关系，所得的正规解可以抑制噪声，但不能安全反映物体的局域性质，文中分析了算法的计算复杂性，指出巨大的计算要求是限制其实用的关键因素。     

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.     

基于远场信息和卷积神经网络的波前重构方法

探测波前相位信息是实现自适应光学波前补偿的关键,使用卷积神经网络（CNN）代替波前传感器进行波前重构,系统简单易于实现,同时重构过程不依赖迭代运算,快速实时。为准确提取远场中的波前特征,CNN需要事先使用大量样本进行训练。研究中根据4～30阶大气湍流泽尼克像差系数与其远场强度的对应关系,仿真制作样本数据集,训练CNN从输入的一帧远场图像中预测出畸变波前的泽尼克像差系数,重构原始波前。验证结果表明,该方法能快速实时地还原出波前相位信息,重构波前较原始波前具有极高的波面吻合度和较小的残差剩余量,有望实现实际自适应光学系统中的闭环校正。     

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     

基于高温超导量子干涉仪的超低场核磁共振成像研究

对基于高温超导量子干涉仪的低场核磁共振成像进行了较为系统的探索.首先对低场核磁共振系统进行了改进和完善,使得装置能够用于成像实验.在此基础上进行了一维、二维成像实验并取得了成功.二维成像分别采用了直接背投影成像法和傅里叶变换重建法.采用直接背投影方法成功获得了不同水样品分布的图形并与实物符合较好,同时还尝试对生物样品如青椒和芹菜的切片进行了成像,也得到了符合原物的二维投影像.尝试用傅里叶变换法对水样品进行成像,得到的图形能够显示样品轮廓,但信噪比偏低.对两种二维成像方法进行了比较和讨论.     

Basic concepts for nuclear magnetic resonance imaging

The basic concepts necessary to understand the physical basis of NMR imaging are presented in this didactic article. It is intended as a starting point for the radiologist or medical physicist who is addressing the topic of NMR for the first time. The basis of the NMR phenomena is described with introduction of the concepts of magnetic moment, magnetic fields, magnetic resonance, net magnetic moment of a sample, NMR excitation and NMR emission. The equipment necessary to observe these NMR properties of matter is summarized as well as the procedures for basic pulsed NMR experiments. The physical concepts for spatial localization of NMR emissions are introduced with physical analogies to stringed musical instruments. Several alternative imaging modalities are compared with greatest emphasis on the inversion recovery technique which yields images weighted by tissue T₁ values. The six subsystems of an NMR imaging device (primary magnet, computer, radio equipment, magnetic gradient, data storage and display subsystems) are described in an overview fashion. The paper is followed by a series of study questions to test the reader's comprehension of basic NMR imaging concepts.     

Flexible reduced field of view magnetic resonance imaging based on single-shot spatiotemporally encoded technique

In many ultrafast imaging applications, the reduced field-of-view(r FOV) technique is often used to enhance the spatial resolution and field inhomogeneity immunity of the images. The stationary-phase characteristic of the spatiotemporallyencoded(SPEN) method offers an inherent applicability to r FOV imaging. In this study, a flexible r FOV imaging method is presented and the superiority of the SPEN approach in r FOV imaging is demonstrated. The proposed method is validated with phantom and in vivo rat experiments, including cardiac imaging and contrast-enhanced perfusion imaging. For comparison, the echo planar imaging(EPI) experiments with orthogonal RF excitation are also performed. The results show that the signal-to-noise ratios of the images acquired by the proposed method can be higher than those obtained with the r FOV EPI. Moreover, the proposed method shows better performance in the cardiac imaging and perfusion imaging of rat kidney, and it can scan one or more regions of interest(ROIs) with high spatial resolution in a single shot. It might be a favorable solution to ultrafast imaging applications in cases with severe susceptibility heterogeneities, such as cardiac imaging and perfusion imaging. Furthermore, it might be promising in applications with separate ROIs, such as mammary and limb imaging.     

基于深度神经网络的镜片佩戴舒适度研究

为了客观评价镜片佩戴舒适度,研究设计眼动实验,利用眼动仪采集并分析眼动数据,提出"眼睑间距变动百分比"作为镜片佩戴舒适度评价指标之一,再结合其他传统眼动指标,设计了一种基于深度神经网络的镜片佩戴舒适度评价模型.结果 表明提出的眼睑间距变动百分比指标用于评价镜片佩戴舒适度具有良好的效果,最终构建的深度神经网络模型预测准确...     

Novel methodology for 3D reconstruction of carotid arteries and plaque characterization based upon magnetic resonance imaging carotid angiography data

In this study, we present a novel methodology that allows reliable segmentation of the magnetic resonance images (MRIs) for accurate fully automated three-dimensional (3D) reconstruction of the carotid arteries and semiautomated characterization of plaque type. Our approach uses active contours to detect the luminal borders in the time-of-flight images and the outer vessel wall borders in the T(1)-weighted images. The methodology incorporates the connecting components theory for the automated identification of the bifurcation region and a knowledge-based algorithm for the accurate characterization of the plaque components. The proposed segmentation method was validated in randomly selected MRI frames analyzed offline by two expert observers. The interobserver variability of the method for the lumen and outer vessel wall was -1.60%±6.70% and 0.56%±6.28%, respectively, while the Williams Index for all metrics was close to unity. The methodology implemented to identify the composition of the plaque was also validated in 591 images acquired from 24 patients. The obtained Cohen's k was 0.68 (0.60-0.76) for lipid plaques, while the time needed to process an MRI sequence for 3D reconstruction was only 30 s. The obtained results indicate that the proposed methodology allows reliable and automated detection of the luminal and vessel wall borders and fast and accurate characterization of plaque type in carotid MRI sequences. These features render the currently presented methodology a useful tool in the clinical and research arena.     

A design of novel type superconducting magnet for super—high field functional magnetic resonance imaging by using the harmonic analysis method of magnetic vector potentials

The approach of expanding the magnetic scalar potential in a series of Legendre polynomials is suitable for designing a conventional superconducting magnetic resonance imaging magnet of distributed solenoidal configuration. Whereas the approach of expanding the magnetic vector potential in associated Legendre harmonics is suitable for designing a single-solenoid magnet that has multiple tiers, in which each tier may have multiple layers with different winding lengths. A set of three equations to suppress some of the lowest higher-order harmonics is found. As an example, a 4T single-solenoid magnetic resonance imaging magnet with 4×6 layers of superconducting wires is designed. The degree of homogeneity in the 0.5m diameter sphere volume is better than 5.8 ppm. The same degree of homogeneity is retained after optimal integralization of turns in each correction layer. The ratio B_m/B₀ in the single-solenoid magnet is 30% lower than that in the conventional six-solenoid magnet. This tolerates higher rated superconducting current in the coil. The Lorentz force of the coil in the single-solenoid system is also much lower than in the six-solenoid system. This novel type of magnet possesses significant advantage over conventional magnets, especially when used as a super-high field functional magnetic resonance imaging magnet.     

磁共振波谱与成像技术

文章简要介绍了核磁共振的基本原理,详细阐述了液体核磁共振在蛋白质结构、功能和动力学等方面的研究进展,论述了增强固体核磁共振分辨率的方法及其应用,讲述了磁共振成像的原理并综述了不同磁共振成像方法的应用研究进展,并对核磁共振的发展作了展望。