基于 GPU 加速技术的 C 形臂 Cone Beam CT 三维 图像快速重建方法

平板探测器技术的发展使得锥形束计算机断层扫描技术(Cone Beam Computerized Tomography,CBCT)成为一种重要的成像技术,有着十分广泛的应用.基于C形臂的CBCT,除了具有CBCT的技术优势外,还特别适合在影像引导介入手术中应用.然而,如何在满足手术实时性要求的同时获得高分辨率高质量的三维断层图像,仍是个十分具有挑战性的课题.文章提出一种基于GPU加速技术的C形臂CBCT三维图像快速重建方法:在算法层面应用GPU并行加速技术对重建算法进行优化,在系统层面通过设计分布式系统和延迟隐藏机制,大大提升了由二维投影图像重建三维体数据的效率.在保持重建精度的前提下,优化后的GPU加速的FDK算法极大地提升了重建过程的计算效率.延迟隐藏机制进一步提升了系统的运行效率.在使用90帧投影时,系统效率提升了26%,重建延迟加速了2.1倍;当使用120帧投影时,系统效率提升39%,重建延迟加速达到3.3倍.     

基于卷积神经网络的欠采样脑部核磁共振图像重建方法

针对目前基于深度学习的欠采样磁共振（MR）图像重建方法都是基于单个切片的重建而忽略相邻切片间的数据冗余的问题,提出一种用于欠采样的多切片脑部MR图像重建的混合级联卷积神经网络（HC-CNN）。首先,将传统的重建方法拓展为基于深度学习的重建模型,并使用级联卷积神经网络来代替传统的迭代重建框架。然后,在每次迭代重建中,分别使用3D卷积模块和2D卷积模块来学习脑部MR图像序列中存在的相邻切片间与单幅切片内部的数据冗余。最后,在每次迭代中使用数据一致性（DC）模块来保持重建图像在k-空间的数据保真度。在单线圈脑部MR图像数据集上的仿真实验结果显示,相较于基于单幅MR图像的重建方法,所提方法在4倍加速因子下的峰值信噪比（PSNR）值平均提升了1.75 dB,在6倍降采样因子下的PSNR值平均提升了2.57 dB,而且该方法的单张图像重建平均用时为15.4 ms。实验结果表明：所提方法不仅能够有效利用切片间的数据冗余并重建出更高质量的图像,而且具有较高的实时性。     

基于卷积神经网络的欠采样脑部核磁共振图像重建方法

针对目前基于深度学习的欠采样磁共振（MR）图像重建方法都是基于单个切片的重建而忽略相邻切片间的数据冗余的问题，提出一种用于欠采样的多切片脑部MR图像重建的混合级联卷积神经网络（HC-CNN）。首先，将传统的重建方法拓展为基于深度学习的重建模型，并使用级联卷积神经网络来代替传统的迭代重建框架。然后，在每次迭代重建中，分别使用3D卷积模块和2D卷积模块来学习脑部MR图像序列中存在的相邻切片间与单幅切片内部的数据冗余。最后，在每次迭代中使用数据一致性（DC）模块来保持重建图像在k-空间的数据保真度。在单线圈脑部MR图像数据集上的仿真实验结果显示，相较于基于单幅MR图像的重建方法，所提方法在4倍加速因子下的峰值信噪比（PSNR）值平均提升了1.75 dB，在6倍降采样因子下的PSNR值平均提升了2.57 dB，而且该方法的单张图像重建平均用时为15.4 ms。实验结果表明：所提方法不仅能够有效利用切片间的数据冗余并重建出更高质量的图像，而且具有较高的实时性。     

基于最小重构误差的优化局部聚合描述符向量图像检索算法

针对局部聚合描述符向量(VLAD)模型中对特征软量化时权重系数的取值不确定性和特征量化误差较大问题,提出一种具有最小重构误差的权重系数分配算法。该算法以最小化重构误差为标准,将具有最小化重构误差的稀疏编码的编码系数作为软量化VLAD的权重系数。数据库的图像检索测试结果表明,该算法相比主流的VLAD特征编码算法所得图像检索精度可提高10%左右,且有更小的特征重构误差。     

基于运动恢复结构方法的植株三维重建

为了探究基于运动恢复结构(Structure From Motion, SFM)方法的植株三维重建模型的效果,为植物三维重建工作提供研究案例,本文以紫叶鸭跖草(setcreasea pallida)为研究对象,在搭建序列图像获取平台的基础上,选取35幅、75幅、105幅序列图像进行三维重建的对比分析;同时从植株表型参数方面,对植株三维重建模型进行精度评价。结果表明：75幅图像序列的重建效果最好;不同图像序列的模型计算的植株高度相对误差（Relative Error, RE）均小于2.5%,决定系数（coefficient of determination, R2）均大于0.998;不同图像序列的模型提取叶片长和叶片宽的RE均小于2.89%,R2均大于0.958。因此,序列图像的数量与重建模型的效果有关,但二者并非呈正相关关系;序列图像的数量对重建叶片的长与宽的误差影响较小;SFM方法应用于结构比较复杂的植株的三维重建可以取得较好的重建效果。     

计算机视觉测量系统的误差模型分析

基于相机的钻孔模型，研究了相机光学系统产生的4种非线性误差，并在有径向失真和图像中心偏移失真的情况下，用模拟数据对匹配误差和不同量化等级在重建过程中的传播过程和影响作了详细讨论，为视觉测量系统达到高精度的三维重建提供了可靠的误差模型。     

Generalized Low Rank Approximations of Matrices

The problem of computing low rank approximations of matrices is considered. The novel aspect of our approach is that the low rank approximations are on a collection of matrices. We formulate this as an optimization problem, which aims to minimize the reconstruction (approximation) error. To the best of our knowledge, the optimization problem proposed in this paper does not admit a closed form solution. We thus derive an iterative algorithm, namely GLRAM, which stands for the Generalized Low Rank Approximations of Matrices. GLRAM reduces the reconstruction error sequentially, and the resulting approximation is thus improved during successive iterations. Experimental results show that the algorithm converges rapidly. We have conducted extensive experiments on image data to evaluate the effectiveness of the proposed algorithm and compare the computed low rank approximations with those obtained from traditional Singular Value Decomposition (SVD) based methods. The comparison is based on the reconstruction error, misclassification error rate, and computation time. Results show that GLRAM is competitive with SVD for classification, while it has a much lower computation cost. However, GLRAM results in a larger reconstruction error than SVD. To further reduce the reconstruction error, we study the combination of GLRAM and SVD, namely GLRAM + SVD, where SVD is preceded by GLRAM. Results show that when using the same number of reduced dimensions, GLRAM + SVD achieves significant reduction of the reconstruction error as compared to GLRAM, while keeping the computation cost low. Editor: Peter Flach     

用EMD和小波消噪的加速度信号压缩重构新方法

针对噪声破坏加速度信号稀疏性、降低其压缩感知重构算法性能问题,提出了一种用经验模态分解(EMD)和小波分析联合消噪的加速度信号压缩重构新方法.该方法首先采用EMD和小波阈值联合消噪方法对加速度信号消噪处理,保持加速度信号内在稀疏性;然后基于压缩感知理论和加速度信号块结构信息,采用块稀疏贝叶斯学习算法以高概率重构原始加速度信号.采用USC-HAD人体日常行为数据库中的加速度信号验证本文方法的有效性.实验结果表明,本文所提方法的信噪比和均方根误差明显优于未经消噪处理的压缩感知重构算法,能够有效抑制加速度信号噪声,增大加速度信号稀疏度,提高加速度信号压缩重构算法性能.     

Three-dimensional reconstruction of the left ventricle from two angiographic views: an evidence combination approach

Clinical interventional hemodynamic studies quantify the ventricular function from two-dimensional (2-D) X-ray projection images without having enough information of the actual three-dimensional (3-D) shape of this cardiac cavity. This paper reports a left ventricle 3-D reconstruction method from two orthogonal angiographic projections. This investigation is motivated by the lack of information about the actual 3-D shape of the cardiac cavity. The proposed algorithm works in 3-D space and considers the oblique projection geometry associated with the biplane image acquisition equipment. The reconstruction process starts by performing an approximate reconstruction based on the Cylindrical Closure Operation and the Dempster-Shafer theory. This approximate reconstruction is appropriately deformed in order to match the given projections. The deformation procedure is carried out by an iterative process that, by means of the Dempster-Shafer and the fuzzy integral theory, combines the information provided by the projection error and the connectivity between voxels. The performance of the proposed reconstruction method is evaluated by considering first the reconstruction of two 3-D binary databases from two orthogonal synthetized projections, obtaining errors as low as 6.48%. The method is then tested on real data, where two orthogonal preprocessed angiographic images are used for reconstruction. The performance of the technique, in this case, is assessed by means of the projection error, whose average for both views is 7.5%. The reconstruction method is also tested by performing the 3-D reconstruction of a ventriculographic sequence throughout an entire cardiac cycle.     

基于仿射CPD的股骨医学图像配准方法*

股骨医学图像配准是股骨三维重建技术的主要研究内容,通过机器辅助获得配准的参数决定了三维重建的准确性。针对传统ICP算法普遍存在的准确度和鲁棒性的较低问题,提出一种基于仿射CPD的股骨医学图像配准方法,使配准自动化以及精准化。该方法通过边缘提取获得浮动、参考点集;在配准后获取结果参数;通过结果参数将股骨通用模型摆成股骨个体化骨折姿态。实验结果：平移、旋转的平均误差分别小于2.1mm、 1.6;配准失败率仅在10%左右;三维重建后的3D图像真实度较高。结论：本文所述的方法对于图像配准和三维重建是可行的和高效的。     

Motion correction based on an appropriate system matrix for statistical reconstruction of respiratory-correlated PET acquisitions

Respiratory motion correction in positron emission tomography (PET) seeks to incorporate motion information into an image reconstruction algorithm by using the full counting statistics of an acquisition to generate a single, motion-free volume. Here, we present a motion-incorporated ordered subsets expectation maximization (MOSEM) reconstruction based on a device-dedicated tomographic projector in which each matrix element is calculated directly from the voxels’ Cartesian coordinates alone. The motion is corrected by updating this projector as a function of the respiratory level. The performance of the reconstruction method was investigated with three datasets: two simulations of a transaxially or axially moving lesion on a patient acquisition and a third acquisition of a moving sphere. After the 16th sub-iteration, the normalized mean square error (NMSE, with a motionless acquisition as reference) was 0.20 for the non-corrected (ungated) image and 0.01 for the MOSEM image with transaxial motion simulation. Likewise, NMSE was 0.30 for the ungated image and 0.03 for MOSEM image with axial motion simulation. For the phantom, ungated reconstruction yielded an error of 0.78, whereas MOSEM yielded 0.43. The error reduction resulted from enhancement and reduced spreading of the moving uptake. Our results show that MOSEM reconstruction yields motion-corrected images which are similar to motionless reference images.     

Two-step Prediction and Image Deflection for Exact Head Tracking in Virtual Environments

Time lag and image jittering are the main reasons for destroying the feeling of reality in most virtual environments. They cause motion sickness because human beings are simply not used to unrealistically generated feedback. This paper proposes advanced software methods that greatly reduce these unwanted effects: a new improved two-step prediction method for better head tracking, an image deflection technique to reduce remaining error and a new software configuration that allows the full use of available hardware capabilities. Our improved prediction technique reduces prediction error by 50% in comparison to standard methods. The image deflection allows to extend the prediction distance by 100% while keeping the error below a tolerable level.     

侵彻加速度信号零漂的处理方法

针对侵彻加速度信号的零漂特征,采用小波分析法处理零漂数据,提出以小波基函数提取趋势项的准确度和信号重构能力作为最优小波基的选择依据。研究表明：零漂信号的频率范围主要集中在100 Hz以内;bior2.8小波基提取趋势项准确度较高,重构侵彻加速度信号的误差最小。含有零漂的侵彻加速度信号采用bior2.8最优小波基分解出100 Hz以下的低频趋势项,能够有效地消除零漂,得到的加速度一次积分与二次积分分别与实际的侵彻速度和侵彻行程保持了较好的一致性,行程相对误差小于10%。消除零漂后的加速度信号可以为计行程自适应起爆方式提供有效信息。     

基于预测稀疏编码的快速单幅图像超分辨率重建

针对经典的基于稀疏编码的图像超分辨率算法在重建过程中运算量大、计算效率低的缺点,提出一种基于预测稀疏编码的单幅图像超分辨率重建算法。训练阶段,该算法在传统的稀疏编码误差函数基础上叠加编码预测误差项构造目标函数,并采用交替优化过程最小化该目标函数;测试阶段,仅需将输入的低分辨图像块和预先训练得到的低分辨率字典相乘就能预测出重建系数,从而避免了求解稀疏回归问题。实验结果表明,与经典的基于稀疏编码的单幅图像超分辨率算法相比,该算法能够在显著减少重建阶段运算时间的同时几乎完全保留超分辨率视觉效果。     

特征点维度的静态场景三维重建运动目标剔除

基于语义分割的图像掩膜方法常用来解决静态场景三维重建任务中运动物体的干扰问题,然而利用掩膜成功剔除运动物体的同时会产生少量无效特征点.针对此问题,提出一种在特征点维度的运动目标剔除方法,利用卷积神经网络获取运动目标信息,并构建特征点过滤模块,使用运动目标信息过滤更新特征点列表,实现运动目标的完全剔除.通过采用地面图像和航拍图像两种数据集以及DeepLabV3、YOLOv4两种图像处理算法对所提方法进行验证,结果表明特征点维度的三维重建运动目标剔除方法可以完全剔除运动目标,不产生额外的无效特征点,且相较于图像掩膜方法平均缩短13.36%的点云生成时间,减小9.93%的重投影误差.     

基于双幅图像三维重建的不确定性分析

由于得到基于误差传播理论的重建不确定性显式公式并没有直观性,文章在给出三维重建扰动分析模型的基础上,应用多元分析的统计方法研究了图像量化误差、匹配误差、标定误差等对重建精度的影响。采用计算机仿真图像进行实验,向重建模型中输入高斯噪声进行扰动分析,这样有利于对不确定性的评定,仿真实验证明了方法的有效性。最后,对三维重建点的伸展不确定性也进行了可视化。     

基于稀疏贝叶斯学习的图像重建方法

结合稀疏贝叶斯学习(SBL)和可压缩传感理论(CS),给出一种在噪声测量条件下重建可压缩图像的方法.该方法将CS理论中图像重建过程看作一个线性回归问题,而待重建的图像是该回归模型中的未知权值参数;利用SBL方法对权值赋予确定的先验条件概率分布用以限制模型的复杂度,并引入超参数;最大化超参数的边缘对数似然函数求得权值参数的最优估计即待重建图像.该方法同时还给出了权值估计的后验概率密度和误差条,从而获得权值最优值的不确定性测量.实验结果表明,SBL方法可以获得精确重建,并且在相同相对重建误差的条件下,比基追踪(BP)方法需要更少的重建时间,比正交匹配追踪(OMP)需要更少的测量次数.     

基于分段式序列图片集的运动恢复结构

多视图运动恢复结构（Structure from Motion,SFM）是三维重建中相机姿态估计的一种最常用的方法。传统SFM采用增量方式处理图片,算法的时间复杂度是[O(n4)],当图片数量较多时,重建时间很长。此外,由于图片噪声影响,漂移误差将随着图片数量增加不断累加,影响最终的重建质量。添加集束调整（Bundle Adjustment,BA）可以优化重建结果,但是需要花费更长的时间。在现有增量式算法的基础上,提出基于分段式序列图片集的方法,将序列图片集按照相似度划分为小集合,对每个小集合进行并行计算,减少误差累积量和重建时间,最后再用BA进行全局优化。实验结果表明,该方法能在保持一定精度的前提下,有效减少重建时间。     

一种恢复图像自然色彩的重构方法

现代数码相机是通过颜色过滤矩阵在每个像素位置采集一个颜色分量,重构出全彩色数字图像。压缩感知理论证明了该重构是误差有界的,但在实际应用时却隐含着一个问题:重构图像所需的稀疏编码字典是从图像数据库学习出来的,而目前数字图像都是重构出来的,因此存在着从重构的图像学习字典去重构图像的循环悖论。针对这个问题,提出并构建了新的完全采样彩色图像的Sandwich图像数据集,打破了压缩感知理论在应用于图像重构时的循环悖论,使得压缩感知方法能够真正地重建自然彩色图像。Sandwich图像数据集的构建及其训练得到的字典可以应用于如图像超分辨率重构、去噪、修复等领域。深入的图像重建实验表明,使用sandwich图像集训练的字典不论是字典原子特性还是由其重构得到的图像质量均好于基于传统数据集的结果。     

Reconstruction error characterization and control: a samplingtheory approach

Reconstruction is prerequisite whenever a discrete signal needs to be resampled as a result of transformations such as texture mapping, image manipulation, volume slicing, and rendering. We present a new method for the characterization and measurement of reconstruction error in the spatial domain. Our method uses the Classical Shannon's Sampling Theorem as a basis to develop error bounds. We use this formulation to provide, for the first time, an efficient way to guarantee an error bound at every point by varying the size of the reconstruction filter. We go further to support position-adaptive reconstruction and data-adaptive reconstruction which adjusts the filter size to the location of the reconstruction point and to the data values in its vicinity. We demonstrate the effectiveness of our methods with 1D signals, 2D signals (images), and 3D signals (volumes)