融合t-分布随机邻域嵌入与自动谱聚类的脑功能精细分区方法

本文针对目前脑功能分区不够准确的问题，基于静息态功能磁共振数据，提出了一种融合t-分布随机邻域嵌入（t-SNE）与自动谱聚类（ASC）的人脑功能精细分区的算法.首先，基于静息态功能磁共振图像，对需功能划分的脑区与全脑的时间序列作相关分析，得到需划分脑区的功能连接模式；然后，利用t-SNE算法提取高维功能连接模式特征；最后，通过基于本征间隙的ASC算法自动确定聚类数目，并对降维后的脑区特征分类，得到精细划分的脑亚区.模拟种子区域上的实验结果表明，相较谱聚类算法，以及结合主成分分析的谱聚类算法，本文方法对脑功能体素划分更优.进一步将本方法应用到真实人脑的功能分区中，成功地将海马旁回分为左右半球各3个亚区.本研究表明使用t-SNE与ASC融合的算法可提高脑功能分区准确性，是脑功能精细分区、进而构建脑功能图谱的一种有效方法.     

多项式展开配准的多图谱大脑磁共振图像分割北大核心CSCD

全脑区的自动分割对于大脑疾病的诊断和治疗具有重要的临床意义。为了提高分割准确率,提出了一种多项式展开配准的多图谱分割算法。利用线性多项式展开模型,结合仿射变换和非刚体变换将待分割的目标图像与图谱图像进行逐一的配准来获取位移场;利用归一化互信息计算二者的相似度,筛选出与目标图像相似度较高的图谱,并对图谱的标记图像进行位移场映射,得到粗分割结果;采用全局加权投票法将粗分割结果进行融合,得到最终的细分割结果。选取了35例来自MICCAI 2012 Multi-Atlas Labeling挑战和10例来自上海交通大学医学院附属第九人民医院的磁共振T1加权图像对该方法进行验证,最终全脑区的DSC值分别为0.7585和0.7351。实验结果表明,算法具有较高的大脑分割精确度和鲁棒性,有望辅助临床进行大脑相关疾病的诊断和治疗。     

基于似然函数最速下降的红外与可见光图像配准

为了实现红外与可见光图像的自动配准,提出了基于似然函数最速下降迭代的图像配准算法.该算法以图像边缘作为配准点特征,将异源图像配准转化为边缘点集配准.基于点集的高斯混合模型建立了边缘点集配准似然函数,以该函数作为目标函数,仿射变换参量作为优化变量,利用最速下降方法进行最优变换参量求解,从而实现边缘点集配准.同时,将多分辨率金字塔引入迭代配准框架下,实现了高分辨率图像配准的加速.实验结果表明:该算法精度高,运算速度快,可以很好地完成可见光与红外图像的自动配准.     

基于心脏电影磁共振图像的一种新的右心室多图谱分割方法

基于心脏电影磁共振图像的右心室(RV)分割,对心脏疾病的诊疗及预后有着十分重要的意义.右心室结构复杂,传统图像分割方法始终未能达到较高的精度.多图谱方法通过配准和融合来实现RV分割,是近几年RV分割中的主要方法之一.本文提出一种新的右心室多图谱分割方法,能够实现RV的全自动准分割.本文首先采用自适应仿射传播算法获取一系列图谱集,并基于豪斯多夫距离和归一化互信息选择与目标图像最相似的图谱集;然后,依次采用多分辨率的仿射变换和Diffeomorphic demons算法将目标图像配准到最相似图谱集,并将配准得到的形变场应用于标记图像获得粗分割结果;最后,本文采用COLLATE算法融合粗分割结果得到RV轮廓.30例心脏电影磁共振数据被用于回顾性分析.本文算法与放射专家手工分割的RV相比,Dice指标和豪斯多夫距离的平均值分别为0.84,11.46 mm;舒张末期容积,收缩末期容积,射血分数的相关系数和偏差均值分别是0.94,0.90,0.86;2.5113,–3.4783,0.0341.与卷积神经网络相比,本文算法在收缩末期的分割精度更接近手动分割结果.实验结果表明,该方法从有效的图谱选择和基于多分辨率的Diffeomorphic demons算法的多级配准提高了右心室分割的精度,有望应用于临床辅助诊断.     

基于似然函数最速下降的红外与可见光图像配准

为了实现红外与可见光图像的自动配准,提出了基于似然函数最速下降迭代的图像配准算法.该算法以图像边缘作为配准点特征,将异源图像配准转化为边缘点集配准.基于点集的高斯混合模型建立了边缘点集配准似然函数,以该函数作为目标函数,仿射变换参量作为优化变量,利用最速下降方法进行最优变换参量求解,从而实现边缘点集配准.同时,将多分辨...     

基于尺度不变特征变换和区域互信息优化的多源遥感图像配准

为了进一步提高遥感图像配准精度, 提出了尺度不变特征变换(SIFT)结合区域互信息优化的遥感图像配准方法. 首先利用混沌序列的随机性和遍历性, 提出一种混沌量子粒子群优化(CQPSO)算法, 在量子粒子群优化(QPSO)算法迭代陷入早熟收敛时, 采用一种新的机理引入混沌序列, 进化粒子克服早熟. 图像配准算法分为预配准和精配准两个过程. 基于SIFT算法提取特征点, 经匹配和有效地外点排除完成预配准, 然后对匹配特征点坐标进行亚像素级微调, 通过最小二乘法求得一系列匹配参数构造初始粒子群, 最后利用混沌量子粒子群优化区域互信息完成精配准, 得到最优匹配参数. 用一些标准测试函数对所提出的CQPSO和QPSO及粒子群优化(PSO)算法进行了实验比较, 另外, 对SIFT, SIFT结合PSO算法优化区域互信息, SIFT结合QPSO算法优化区域互信息和SIFT结合CQPSO算法优化区域互信息(SRC)等四种算法进行了不同分辨率遥感图像配准实验比较和不同时相遥感图像配准实验比较, 实验结果验证了所提出的CQPSO算法的优越性和SRC配准方法的有效性.     

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.     

一种用于钣金零件视觉测量的图像拼接方法

在钣金零件视觉测量中,大尺寸零件超出图像采集摄像机的视场范围,不能一次性获得该零件的完整图像,需要将存在一定重叠区域的图像进行图像拼接。针对特征稀少、灰度信息少、轮廓对称的钣金零件图像不易拼接问题进行了分析和研究,提出了一种在待拼接图像中选择包含重叠部分的等大小区域作为待配准区域,并对其提取轮廓,然后对轮廓图利用相位相关法求待配准区域的配准参数,再根据待配准区域图像与待拼接图像的坐标关系转换求待拼接图像的配准参数并对图像进行配准,实现钣金零件图像拼接的方法。通过实验表明,该拼接方法对特征稀少、灰度信息少和轮廓对称的钣金零件图像具有速度快、较好的准确性和抗干扰能力,能够满足工业应用的要求。     

基于Fourier-Mellin变换的气象卫星光谱图像配准

气象卫星光谱图像是气象科学和环境遥感科学研究的重要工具,而图像配准是气象卫星图像数据应用的前提。文章针对气象卫星光谱图像的配准问题,提出了一种基于Fourier-Mellin变换的自动配准方法。首先利用全球海岸线矢量图数据构造地标模板,地标模板是气象卫星光谱图像配准的参考图像;其次,根据云通道数据选择无云区域红外子图像,并利用Sobel算子对红外光谱图像提取边缘特征;最后利用Fourier-Mellin变换确定地标模板图像和红外边缘图像之间的仿射变换参数,从而实现红外光谱图像的配准。该方法本质是基于曲线匹配的思想,无需特征点提取,大大简化了配准流程。利用FY-2D气象卫星上获取的红外通道数据进行了实验,结果表明：该方法鲁棒性好,运算速度快,配准精度较高。     

同步辐射纳米CT图像配准方法研究

基于同步辐射的X射线纳米成像技术是无损研究物质内部纳米尺度结构的强大工具,本文总结了图像配准技术在纳米CT成像领域的研究和应用,并根据发展阶段进行分类分析.首先,通过统计近年以来图像配准文献的发表情况,分析并预测纳米尺度图像配准的未来研究方向.其次,基于图像经典配准算法理论,详细介绍了图像配准算法在纳米成像领域最有效的前沿应用.最后,介绍了基于深度学习的图像配准方法的前沿研究,并讨论深度学习在纳米分辨图像配准领域的适用性及发展潜能,根据纳米尺度图像数据的特点及各种深度学习网络模型的特性,展望了同步辐射纳米尺度图像配准技术的未来研究方向及挑战.     

Cellular neural networks for welding arc thermograms segmentation

Machine vision systems are used in many areas for monitoring of technological processes. Among this processes welding takes important place, where often infrared cameras are used. Besides reliable hardware, successful application of vision systems requires suitable software based on proper algorithms. One of most important group of image processing algorithms is connected to image segmentation. Obtainment of exact boundary of an object that changes shape in time, such as the welding arc, represented on a thermogram is not a trivial task. In the paper a segmentation method using supervised approach based on a cellular neural networks is presented. Simulated annealing and genetic algorithm were used for training of the network (template optimization). Comparison of proposed method to a well elaborated segmentation method based on region growing approach was made. Obtained results prove that the cellular neural network can be a valuable tool for infrared welding pool images segmentation.     

An improved FSL-FIRST pipeline for subcortical gray matter segmentation to study abnormal brain anatomy using quantitative susceptibility mapping (QSM)

Accurate and robust segmentation of subcortical gray matter (SGM) nuclei is required in many neuroimaging applications. FMRIB's Integrated Registration and Segmentation Tool (FIRST) is one of the most popular software tools for automated subcortical segmentation based on T₁-weighted (T1w) images. In this work, we demonstrate that FIRST tends to produce inaccurate SGM segmentation results in the case of abnormal brain anatomy, such as present in atrophied brains, due to a poor spatial match of the subcortical structures with the training data in the MNI space as well as due to insufficient contrast of SGM structures on T1w images. Consequently, such deviations from the average brain anatomy may introduce analysis bias in clinical studies, which may not always be obvious and potentially remain unidentified. To improve the segmentation of subcortical nuclei, we propose to use FIRST in combination with a special Hybrid image Contrast (HC) and Non-Linear (nl) registration module (HC-nlFIRST), where the hybrid image contrast is derived from T1w images and magnetic susceptibility maps to create subcortical contrast that is similar to that in the Montreal Neurological Institute (MNI) template. In our approach, a nonlinear registration replaces FIRST's default linear registration, yielding a more accurate alignment of the input data to the MNI template. We evaluated our method on 82 subjects with particularly abnormal brain anatomy, selected from a database of > 2000 clinical cases. Qualitative and quantitative analyses revealed that HC-nlFIRST provides improved segmentation compared to the default FIRST method.     

Rat brain digital stereotaxic white matter atlas with fine tract delineation in Paxinos space and its automated applications in DTI data analysis

PurposeTo automatically analyze diffusion tensor images of the rat brain via both voxel-based and ROI-based approaches, we constructed a new white matter atlas of the rat brain with fine tracts delineation in the Paxinos and Watson space.Materials and methodsUnlike in previous studies, we constructed a digital atlas image from the latest edition of the Paxinos and Watson. This atlas contains 111 carefully delineated white matter fibers. A white matter network of rat brain based on anatomy was constructed by locating the intersection of all these tracts and recording the nuclei on the pathway of each white matter tract. Moreover, a compatible rat brain template from DTI images was created and standardized into the atlas space. To evaluate the automated application of the atlas in DTI data analysis, a group of rats with right-side middle cerebral artery occlusion (MCAO) and those without were enrolled in this study.ResultsThe voxel-based analysis result shows that the brain region showing significant declines in signal in the MCAO rats was consistent with the occlusion position.ConclusionWe constructed a stereotaxic white matter atlas of the rat brain with fine tract delineation and a compatible template for the data analysis of DTI images of the rat brain.     

结合小波融合和深度学习的脑胶质瘤自动分割

针对水肿区域边界模糊和瘤内结构复杂多变导致的脑胶质瘤分割不精确问题,本文提出了一种基于小波融合和3D-UNet网络的脑胶质瘤磁共振图像自动分割算法.首先,对脑胶质瘤磁共振图像的T1、T1ce、T2、Flair四种模态进行小波融合以及偏置场校正;然后,提取待分类的图像块;再利用提取的图像块训练3D-UNet网络以对图像块中的像素进行分类;最后加载损失率较小的网络模型进行分割,并采用基于连通区域的轮廓提取方法,以降低假阳性率.对57组Brats2018（Brain Tumor Segmentation 2018）磁共振图像测试集进行分割的结果显示,肿瘤的整体、核心和水肿部分的平均分割准确率（DSC）分别达到90.64%、80.74%和86.37%,这表明该算法分割脑胶质瘤准确率较高,与金标准相近.相比多模态图像融合前,该算法在减少输入网络数据量和图像冗余信息的同时,还一定程度上解决了胶质瘤边界模糊、分割不精确的问题,提高了分割的准确度和鲁棒性.     

BRAHMA: Population specific T1, T2, and FLAIR weighted brain templates and their impact in structural and functional imaging studies

Differences in brain morphology across population groups necessitate creation of population-specific Magnetic Resonance Imaging (MRI) brain templates for interpretation of neuroimaging data. Variations in the neuroanatomy in a genetically heterogeneous population make the development of a population-specific brain template for the Indian subcontinent imperative. A dataset of high-resolution 3D T1, T2-weighted, and FLAIR images acquired from a group of 113 volunteers (M/F - 56/57, mean age-28.96 ± 7.80 years) are used to construct T1, T2-weighted, and FLAIR templates, collectively referred to as Indian Brain Template, “BRAHMA”. A processing pipeline is developed and implemented in a MATLAB based toolbox for template construction and generation of tissue probability maps and segmentation atlases, with additional labels for deep brain regions such as the Substantia Nigra generated from the T2-weighted and FLAIR templates. The use of BRAHMA template for analysis of structural and functional neuroimaging data obtained from Indian participants, provides improved accuracy with statistically significant results over that obtained using the ICBM-152 (International Consortium for Brain Mapping) template. Our results indicate that segmentations generated on structural images are closer in volume to those obtained from registration to the BRAHMA template than to the ICBM-152. Furthermore, functional MRI data obtained for Working Memory and Finger Tapping paradigms processed using the BRAHMA template show a significantly higher percentage of the activation area than ICBM-152 in relevant brain regions, i.e. the left middle frontal gyrus, and the left and right precentral gyri, respectively. The availability of different image contrasts, tissue maps, and segmentation atlases makes the BRAHMA template a comprehensive tool for multi-modal image analysis in laboratory and clinical settings.     

Segmentation of acoustic images by neural network processing

A segmentation method for biomedical acoustic images is reported which efficiently classifies the groups of similar image elements (pixels) and separates them into particular characteristic regions. As the input data, the method uses the pixel intensities of the source image. The classification is performed by learning vector quantization neural networks, which separate the main classes (structures, tissues, artifacts, etc.) present in the image. Because this type of neural network implies that the number of the classes is known and that the network should be trained by instruction, an expert must participate in the process of generating the input data. Results obtained by processing test acoustic (ultrasonic) images demonstrate that the method is capable of effectively solving sonography classification problems. The accuracy of the method is estimated by comparison with the segmentation performed manually.     

基于多层神经网络的非线性图像分割

提出了一种用多层神经网络对图像进行非线性分割的方法。讨论了所用多层神经网络的学习速度的改进与训练样本的选择方法。实验表明，该多层神经网络系统可用于实时图像分割，并能获得很好的结果。     

基于热红外图像处理技术的农作物冠层识别方法研究

为解决农作物冠层热红外图像边缘灰度级分布不均且噪声较大,而传统图像分割方法难以实现其目标区域有效识别的难题,以苗期红小豆冠层热红外图像为研究对象,将模糊神经网络和仿射变换有机结合,提出了基于热红外图像处理技术的农作物冠层识别模型。首先利用五层线性归一化模糊神经网络的自适应特性,选取高斯隶属度函数,自动计算冠层可见光图像识别的推理规则,有效地分割了可见光图像中的冠层区域。通过分析3种分割指标和熵,定量评价可见光图像冠层分割质量。网络迭代38次时,误差精度为0.000 952,该算法平均有效识别率为96.13%,获取可见光冠层图像的像元信息熵值范围为2.454 4~5.198 7,与标准算法所得冠层图像的像元信息熵仅相差0.245 9。然后以取得可见光图像的冠层有效区域为参考图像,采用仿射变换算法,调整优选平移、旋转、缩放等图像变换因子,配准原始热红外图像,提出了基于仿射变换的冠层热红外图像识别方法。对于初始温度范围值在16.35~19.92 ℃的农作物热红外图像,计算选取旋转幅度为1.0和缩放因子为0.9时,作为异源图像的最优配准参数,获取目标图像的最大温差为3.17 ℃,相对于原图像的平均温度值由18.711 ℃下降至17.790 ℃,进而实现了基于热红外图像处理技术的农作物冠层识别。最后以熵的互信息作为监督指标,对农作物冠层热红外图像识别方法进行评价。提出的冠层热红外图像识别方法,所获取的目标图像与初始热红外图像的平均互信息为4.368 7,标准目标图像和初始热红外图像的平均互信息为3.981 8,二者仅相差0.486 9。同时,两种冠层热红外图像的平均温度差值为0.25 ℃,高效消除了原始热红外图像的背景噪声。结果表明本研究方法的有效性和实用性,能够为应用热红外图像反映农作物生理生态信息特征指标参数提供技术借鉴。     

基于最小平均距离免疫算法的模糊红外图像分割

由于犯罪分子利用各种方法来避开传统的刑侦图像技术,因而红外图像逐渐成为获取犯罪现场痕迹的有效手段。然而,从犯罪现场拍摄的红外图像其目标痕迹大多是弱化的,所以在这类红外图像中分割目标是一项具有挑战性的任务。已有基于生物免疫的各类算法尚未明确描述免疫分割作用领域,以及免疫网络算法模型中的免疫识别距离。为实现对目标痕迹弱化红外图像的有效分割,提出了一种新的具有免疫作用领域和最小平均免疫识别距离的人工免疫构架,设计了一种具备最小平均距离免疫域的免疫分割算法。该方法根据红外图像的特点,采用多步分类算法、免疫变异和自适应免疫最小均距识别方法,根据目标区域和背景区域的总体统计特性实现最佳分类。实验结果表明,提出的基于最小平均距离的免疫算法能够有效地分割目标弱化的红外图像。与经典的边缘模板和区域模板方法相比,该算法具有更好的分割效果,尤其是针对目标弱化红外图像的分割,该算法能够较好地给出五个手指的边界轮廓。     

A New Pooling Approach Based on Zeckendorf’s Theorem for Texture Transfer Information

The pooling layer is at the heart of every convolutional neural network (CNN) contributing to the invariance of data variation. This paper proposes a pooling method based on Zeckendorf’s number series. The maximum pooling layers are replaced with Z pooling layer, which capture texels from input images, convolution layers, etc. It is shown that Z pooling properties are better adapted to segmentation tasks than other pooling functions. The method was evaluated on a traditional image segmentation task and on a dense labeling task carried out with a series of deep learning architectures in which the usual maximum pooling layers were altered to use the proposed pooling mechanism. Not only does it arbitrarily increase the receptive field in a parameterless fashion but it can better tolerate rotations since the pooling layers are independent of the geometric arrangement or sizes of the image regions. Different combinations of pooling operations produce images capable of emphasizing low/high frequencies, extract ultrametric contours, etc.