Non-rigid landmark-based large-scale image registration in 3-D reconstruction of mouse and rat kidney nephrons

BackgroundSerial histological sections are suffering from mechanical distortions that disturb the reconstruction of 3-D objects. We have corrected such artifacts with a non-rigid landmark-based method that respects the original geometry in the tissue block. The method is exemplified on a large scale in the registration of semi-thin serial sections of the mouse and rat kidneys, and has been tested on FFPE-sections.AimIn this study of mouse and rat kidneys, we have measured and characterized the deformations introduced in the preparation of 2.5-μm-thick Epon sections and then eliminated them by a landmark-based non-rigid transformation (NRT).MethodsWe obtained 2.5-μm-thick serial Epon sections from three mouse kidneys and three rat kidneys for 3-D reconstruction of the nephron tubules. First, the images from 3000 serial mouse and 13,000 serial rat sections underwent a classic rigid registration (CRR), and the distortions were measured and indexed. The section images underwent a further NRT in order to compensate for the deformations. The NRT used is a classic interactive landmark-based approach. The quality of the NRT was verified by comparing the geometry of the transformed images with corresponding block images.ResultsAfter CRR, the 2.5-μm-thick sections had a linear deformation of up to 2%, the tubular lengths were overestimated with up to 1.5×, and it was most difficult to trace the tubules from section to section. After the additional NRT, the geometry of the images reflected the original geometry in the block, the tubular lengths were no longer overestimated, and the NRT highly facilitated the tracing of the tubular system.ConclusionsNRT has facilitated the tracing of the tubular system in kidneys, a tracing, which would otherwise have been most difficult to perform. NRT has yielded substantial new knowledge to segmental and spatial nephron organization in the mouse and rat kidneys.     

Serial sectioning and three-dimensional reconstruction of mouse Peyer's patch

Peyer's patches (PPs) are typical gut-associated lymphoid tissues that are located along the wall of the small intestine and that serve as the major sites for generation of immunity to intestinal antigens. Their unique micro-organization is crucial for the generation of the immune response. Although many studies have been reported on the functional anatomy of PP, most investigations have relied on the random sampling of these organs, a procedure that is insufficient for the systemic scanning of the whole tissue or organ. By combining a variety of methods, we have accomplished 3D reconstructions of Peyer's patch. The complex reconstruction procedure includes several steps. First, the PP are serially sectioned at a thickness of 10 microm with a cryostat; (b) the serial sections are stained with haematoxylin-eosin; (c) multiple images from the PP are acquired with an automatic microscope and stitched together with Image Pro Plus to generate a composite image for the whole organ; (d) the serial images are reconstructed with Image J, Reconstruct and 3D Studio Max. The combinational approaches that we present here should be of value when extrapolated to the reconstruction of other tissues or organs. Moreover, the 3D model that we have created and our stereological analysis should be extremely helpful for further in vivo microscopic studies of PP with respect to the immune response.     

A novel algorithm for automatic localization of human eyes

Based on geometrical facial features and image segmentation, we present a novel algorithm for automatic localization of human eyes in grayscale or color still images with complex background. Firstly, a determination criterion of eye location is established by the prior knowledge of geometrical facial features. Secondly, a range of threshold values that would separate eye blocks from others in a segmented face image (i.e., a binary image) are estimated. Thirdly, with the progressive increase of the threshold by an appropriate step in that range, once two eye blocks appear from the segmented image, they will be detected by the determination criterion of eye location. Finally, the 2D correlation coefficient is used as a symmetry similarity measure to check the factuality of the two detected eyes. To avoid the background interference, skin color segmentation can be applied in order to enhance the accuracy of eye detection. The experimental results demonstrate the high efficiency of the algorithm and correct     

一种目标级的遥感图像变化检测算法

传统的像素级变化检测方法对图像的配准准确度要求较高,因而在实际运用中受到很多限制.在人造目标检测的基础上,提出了一种目标级的基于局部配准误差补偿的变化检测方法.根据遥感图像中人造目标与自然目标的纹理差异,对图像中的人造目标进行检测和分割,再对分割图像采用提出的算法进行变化检测.实验表明,与传统的像素级变化检测方法相比,本算法具有较高的检测准确度,对配准准确度的要求也有所放宽,并且可以简化变化检测前的辐射校正工作和变化检测后的像素分类的工作.     

Dual-mode registration of dynamic contrast-enhanced ultrasound combining tissue and contrast sequences

This study proposes a new method for automatic, iterative image registration in the context of dynamic contrast-enhanced ultrasound (DCE-US) imaging. By constructing a cost function of image registration using a combination of the tissue and contrast-microbubble responses, this new method, referred to as dual-mode registration, performs alignment based on both tissue and vascular structures. Data from five focal liver lesions (FLLs) were used for the evaluation. Automatic registration based on the dual-mode registration technique and tissue-mode registration obtained using the linear response image sequence alone were compared to manual alignment of the sequence by an expert. Comparison of the maximum distance between the transformations applied by the automatic registration techniques and those from expert manual registration reference showed that the dual-mode registration provided better precision than the tissue-mode registration for all cases. The reduction of maximum distance ranged from 0.25 to 9.3 mm. Dual-mode registration is also significantly better than tissue-mode registration for the five sequences with p  -values lower than 0.03$0.03$. The improved sequence alignment is also demonstrated visually by comparison of images from the sequences and the video playbacks of the motion-corrected sequences. This new registration technique better maintains a selected region of interest (ROI) within a fixed position of the image plane throughout the DCE-US sequence. This should reduce motion-related variability of the echo-power estimations and, thus, contribute to more robust perfusion quantification with DCE-US.     

Accuracy of image registration between MRI and light microscopy in the ex vivo brain

A multistep procedure was developed to register magnetic resonance imaging (MRI) and histological data from the same sample in the light microscopy image space, with the ultimate goal of allowing quantitative comparisons of the two datasets. The fixed brain of an owl monkey was used to develop and test the procedure. In addition to the MRI and histological data, photographic images of the brain tissue block acquired during sectioning were assembled into a blockface volume to provide an intermediate step for the overall registration process. The MR volume was first registered to the blockface volume using a combination of linear and nonlinear registration, and two dimensional (2D) blockface sections were registered to corresponding myelin-stained sections using a combination of linear and nonlinear registration. Before this 2D registration, two major types of tissue distortions were corrected: tissue tearing and independent movement of different parts of the brain, both introduced during histological processing of the sections. The correction procedure utilized a 2D method to close tissue tears and a multiple iterative closest point (ICP) algorithm to reposition separate pieces of tissue in the image. The accuracy of the overall MR to micrograph registration procedure was assessed by measuring the distance between registered landmarks chosen in the MR image space and the corresponding landmarks chosen in the micrograph space. The average error distance of the MR data registered to micrograph data was 0.324±0.277 mm, only 8% larger than the width of the MRI voxel (0.3 mm).     

An improved image alignment procedure for high-resolution transmission electron microscopy

Image alignment is essential for image processing methods such as through-focus exit-wavefunction reconstruction and image averaging in high-resolution transmission electron microscopy. Relative image displacements exist in any experimentally recorded image series due to the specimen drifts and image shifts, hence image alignment for correcting the image displacements has to be done prior to any further image processing. The image displacement between two successive images is determined by the correlation function of the two relatively shifted images. Here it is shown that more accurate image alignment can be achieved by using an appropriate aperture to filter the high-frequency components of the images being aligned, especially for a crystalline specimen with little non-periodic information. For the image series of crystalline specimens with little amorphous, the radius of the filter aperture should be as small as possible, so long as it covers the innermost lattice reflections. Testing with an experimental through-focus series of Si[1 1 0] images, the accuracies of image alignment with different correlation functions are compared with respect to the error functions in through-focus exit-wavefunction reconstruction based on the maximum-likelihood method. Testing with image averaging over noisy experimental images from graphene and carbon-nanotube samples, clear and sharp crystal lattice fringes are recovered after applying optimal image alignment.     

基于角点的红外与可见光图像自动配准方法

针对红外图像与可见光图像的自动配准问题,提出了一种基于图像角点特征以及仿射变换模型的方法.利用Harris因子分别在红外图像和可见光图像上检测角点,并对两幅图像进行边缘检测,得到其边缘图像.通过角点邻域在边缘图像上的相关性,实现角点的粗匹配;通过角点的细匹配,从匹配的角点中选择两对匹配最佳的点作为仿射变换的控制点,得到仿射变换模型,并对待配准图像进行仿射变换,从而实现图像配准.实验结果表明:该方法运算速度快,可以很好地完成红外与可见光图像的自动配准.     

An optimization method for registration and mosaicking of remote sensing images

An optimization method for registration and mosaicking of remote sensing images has been proposed. More than four pairs of control points extracted by scale invariant feature transform (SIFT) detector have been used to obtain the projective transformation relation between the reference image and text image. A method based on gradient information of the region has been proposed to assign the weighted value to the feature points. The weighted value reflects the importance of the area, which can change the result of optimization. The combination method of seam cutting and feathering technique is used to stitch two overlapped images. Our experiment shows that the alignment error of interest area is smaller than that of normal area. If the weighted parameter is given proper value, the alignment error of interest area can reduced to less than one pixel.     

Super-resolution reconstruction face recognition based on multi-level FFD registration

Super-resolution (SR) reconstruction is an effective method to solve the problem, that the face image resolution is too low to be recognized in video, but the non-rigid change of deformed face and expression changes greatly affect the accuracy of registration and reconstruction. To solve these problems, a method of multi-level model free form deformation (FFD) elastic registration algorithm based on B spline is proposed. It first use low-resolution FFD grid for global registration, to emphasize the contribution of edge information for registration, we introduce edge registration measure into the sum of squared difference (SSD) criterion. Then divide the global registration image and reference image into a series of corresponding sub-image pairs and calculate the correlation coefficient of each pair; at the same time, we do local registration with high-resolution FFD grid to the small value correlation coefficient sub-image pairs. In the registration process of optimization, the paper uses adaptive step length gradient descent method algorithm based on chaotic variables to improve optimization efficiency. After registration, the algorithm of project onto convex sets (POCS) is used to reconstruct SR face image through several low resolution image sequences, and then recognized these SR face images by support vector machines (SVM) classifier. Experimental results from standard video database and self-built video database show that this method can register and reconstruct face image accurately in the condition of great face deformation and expression change, while the face recognition accuracy is also improved.     

Quantifying the architectural complexity of microscopic images of histology specimens

Tumour grade (a measure of the degree of cellular differentiation of malignant neoplasm) is an important prognostic factor in many types of cancer. In general, poorly differentiated tumours are characterized by a higher degree of architectural irregularity and complexity of histological structures. Fractal dimension is a useful parameter for characterizing complex irregular structures. However, one of the difficulties of estimating the fractal dimension from microscopic images is the segmentation of pathologically relevant structures for analysis. A commonly used technique to segment structures of interest is to apply a pixel intensity threshold to convert the original image to binary and extract pixel outline structures from the binary representation. The difficulty with this approach is that the value of the threshold required to segment the histological structures is highly dependent on the staining technique chosen and imaging conditions (i.e., illumination time, intensity, and uniformity) of the microscopic system. In this work, we present a method for finding the optimal intensity threshold by maximizing the corresponding fractal dimension. This method results in the segmentation of histological structures and the estimation of their fractal dimension (independent of imaging conditions). We applied our technique to 164 prostate histology sections from 82 prostate core biopsy specimens (two serial sections from each of the 63 benign prostate tissues and 19 high grade prostate carcinoma). We stained one of the serial sections with conventional hemotoxylin and eosin (H&E) and the other with pan-keratin, and found that the difference in mean fractal dimension between the two groups was statistically significant (p < 0.0001) for both stains. However, using receiver operating characteristics (ROC) analysis, we conclude that our fractal dimension method applied to the images of pan-keratin stained sections provides greater classification performance (benign versus high grade) than with those stained with H&E when compared to the original histological diagnosis. The sensitivity and specificity achieved with the pan-keratin images were 89.5% and 90.5%, respectively.     

Image registration framework for large-scale longitudinal MRI data sets: strategy and validation

Advanced magnetic resonance imaging (MRI) studies often require the transformation of large numbers of images into a common space. Calculating transformations that relate each image to every other and applying them to the images on demand are theoretically possible; however, these can be computationally prohibitive. Therefore, relating each image to only one other image, then linking those transforms together to relate any two images in the database, may be an efficient alternative. Evaluated were the feasibility and validity of image registration to bring intraindividual MR images into mutual correspondence for longitudinal analysis through the concatenation of precomputed transforms. A longitudinal data set of 10 multiple sclerosis patients with nine serial dual-echo spin-echo, 1.5-T MRI scans was used. Intrasubject registrations were performed stepwise between consecutive images and direct from each time point to the baseline. Consecutive transforms were concatenated and evaluated against direct registrations by comparing the resulting transformed images (using Pearson correlation coefficient). Confounding variables such as time between scans, brain atrophy, and change in lesion load were evaluated. We found the images resampled with the direct and the concatenated transforms to be highly correlated, and there was no significant difference between methods. Differences in brain parenchymal fraction (a measure of brain atrophy) showed significant inverse correlation with the correspondence of the resampled images. Results indicate that concatenating multiple transforms that link two images together produces near-identical results to that of direct registration; thus, this method is both useful and valid.     

快速消除车辆阴影的多阈值图像分割法

 视频图像分割时的运动阴影由于与被测对象的相似性而常被误判为被测对象,传统的阴影检测方法一般难以满足实时智能交通系统对处理速度的要求。为此,提出用多阈值法分割图像,将灰度化的当前图像与背景差分,再用正、负两个阈值对其二值化,在分割出深色和浅色被测对象的同时消除阴影。实验表明,将这种方法应用于运动车辆的检测,具有速度快和阴影消除效果好等特点,可应用于实时的运动目标检测和跟踪等领域。     

基于投影的X光图像分割区域数量判定

为了在随身行李X光图像中自动地检测和识别低亮度的危险物品,首先需要将原始图像进行分割,以得到各个目标的区域.分割图像既不能过分割,也不能欠分割.本文提出了一种基于投影的X光图像分割区域数量判定方法,通过比较分割图像序列在水平和垂直方向投影的相关程度的变化,确定最佳的图像分割区域数量.和基于统计有效指数的方法相比,该方法综合考虑了图像的统计信息和空间分布信息,不仅所得到的最优分割图像中包含足够的目标信息,便于后续的目标检测和识别,而且算法具有较小的计算量.     

高功率激光器自动准直系统图像处理的可信度评估

自动准直系统是将采集的图像进行特定算法处理获取光束的位置信息,驱动电机调整光斑到指定的位置,其中图像质量对光束定位的精确性影响很大,为尽量避免受严重噪音和光束畸变影响的图像进入图像处理流程而产生准直结果的错误判断,保证自动准直系统图像处理结果的精确性,试图提供一种较为简单的判断依据,采用基于Monte-Carlo模拟方法,建立了远场准直过程中三种主要噪音的干扰评估模型,以测量不确定度作为定量判断处理图像与否的特征参数,将该结果做成统计图表作为判断依据。结果表明,在设定合理测量不确定度阈值的情况下,该特征参数能够较好地排除不良图像,从而提高准直结果的可信度。     

基于图像局部区域梯度特征描述的红外人体识别算法

提出了一种新的红外图像中人体目标识别方案并进行了算法实现。通过直方图聚类分析对红外图像进行分割,根据二值化图像团块的特点,确定图像中的候选目标图像区域。将候选目标图像按比例划分为多个区域,使用梯度位置朝向直方图(GLOH,Gradient location-orientation histogram)对候选目标图像进行描述。与其它红外图像中人体识别算法相比,不需要多种特征提取算法组合进行分步骤识别,仅使用单个SVM分类器即可达到满意的识别率,避免了分类器的级联,算法简单有效。     

Application of color image segmentation to estrusc detection

Automatic segmentation and classification of color images is a problem of great practical interest in different areas. This paper presents an algorithm for this purpose which is divided in three steps. Firstly, the regions of interest are isolated from the rest of the image based on threshold functions defined in theYUV andYIQ color spaces, producing a set of connected components. Then, a set of features is computed to enable a quantitative evaluation of the segmented objects. Finally, the image is classified by means of a decision rule based on the analysis of the differences between the computed measures and a set of ideally segmented images, according to experts’ assessment. The algorithm was applied to a decision support tool for estrus detection in cattle. This approach constitutes a valuable alternative to improve this process, as it may replace the visual observation by the automatic analysis of pictures taken to cows in controlled environments. Experimental results show that the segmentations obtained with this method are highly satisfactory and they allow a precise classification of the images with low computational complexity.     

Performance evaluation of maximal separation techniques in immunohistochemical scoring of tissue images

This paper presents an automatic scoring method for p53 immunostained tissue images of oral cancer that consist of tissue image segmentation, splitting of clustered nuclei, feature extraction and classification. The tissue images are segmented using entropy thresholding technique in which the optimum threshold value to each color component is obtained by maximizing the global entropy of its gray-level co-occurrence matrix and clustered cells are separated by selectively applying marker-controlled watershed transform. Cell nuclei feature is extracted by maximal separation technique (MS) based on blue component of tissue image and subsequently, each cell is classified into one of four categories using multi-level thresholding. Finally, IHC score of tissue images have been determined using Allred method. A statistical analysis is performed between immuno-score of manual and automatic method, and compared with the scores that have obtained using other MS techniques. According to the performance evaluation, IHC score based on blue component that has high correlation coefficients (CC) of 0.95, low mean difference (MD) of 0.15, and a very close range of 95% confidence interval with manual scores. Therefore, automatic scoring method presented in this paper has high potential to help the pathologist in IHC scoring of tissue images.     

基于相位一致性和Hough变换的多源图像配准方法

由于红外图像与可见光图像对比度不同,常用基于梯度幅值的特征匹配方法难以正确配准。在分析红外图像与可见光图像成像机制的基础上,提出了一种结合相位一致性边缘检测与Hough变换的多源图像配准新方法。该算法首先采用高通滤波和平台直方图均衡方法对红外图像进行预处理以提高红外图像的对比度,再利用具有图像对比度不变性的相位一致性边缘检测法提取两幅图像的边缘,结合Hough变换选取图像空间中最长的线作为特征,采用改进相位相关法作为相似性度量,在对数极坐标域下计算出两幅图像的几何变形参数。仿真实验结果表明,该方法能够以较高查准率实现红外与可见光图像自动配准,并具有较强的鲁棒性。     

基于边缘与SURF算子的SAR与可见光图像配准方法

鉴于SAR(synthetic aperture radar)与可见光图像的成像机理存在很大差别,使得其同名特征的提取和配准十分困难,但在某些情况下,这两类图像的边缘存在一定的相关性。提出一种基于边缘与SURF(speed-up robust feature)算子的图像配准方法。通过适当预处理增强图像间的共性,采用综合性能比较好的Canny算子提取两幅图像共有的边缘特征,在边缘图像的基础上提取SURF特征;通过比值提纯法进行特征点粗匹配,RANSAC(random sample consensus)算法剔除误匹配点,计算仿射变换模型从而实现SAR与可见光图像的自动配准。实验结果表明:该算法的正确匹配率为100%,均方根误差为0.852个像素,配准精度达到亚像素水平。