基于配准前特征提取的颅脑病变检出方法研究

针对从高分辨率颅脑CT图像中自动检出病变以实现计算机辅助诊断的需求,提出配准前特征提取的方法。该方法的主要特点是在图谱创建过程中使用散布点内插法获得整数点的特征值,而在病变检出时则采用格子点内插法获得非整数点的特征值。相对于配准后的特征提取,配准前的特征提取能够更加准确地描述图像的灰度特征和纹理特征。通过实验验证,基于配准前特征提取的颅脑病变检出方法能够提高病变检出率,提高检测精度,但同时也增加了假阳性率。为了减少假阳性率、进一步提高检测精度,在下一步的工作中要研究基于三维体数据的病变检出方法,同时还需要进一步研究非刚性配准的可逆性。     

基于灰度和形状的非刚性图像配准算法的研究

提出一种新的灰度和形状信息相结合的全自动同模态医学图像非刚性配准-分割算法,将欧氏距离表示的形状信息融入基于灰度的配准算法中,构造出新的代价函数.该算法在医学图像多目标分割的应用中,能够较好地完成灰度相近、边缘模糊、间距较小的不同结构的分割.对5组真实脑部MRI图像进行分割脑深层灰质结构的实验,结果表明,本算法优于基于灰度信息的图像配准算法.     

基于图谱的肝脏CT三维自动分割研究

目的在肝脏外科手术或肝脏病理研究中,计算肝脏体积是重要步骤。由于肝脏外形复杂、临近组织灰度值与之接近等特点,肝脏的自动医学图像分割仍是医学图像处理中的难点之一。方法本文采用图谱结合3D非刚性配准的方法,同时加入肝脏区域搜索算法,实现了鲁棒性较高的肝脏自动分割程序。首先,利用20套训练图像创建图谱,然后程序自动搜索肝脏区域,最后将图谱与待分割CT图像依次进行仿射配准和B样条配准。配准以后的图谱肝脏轮廓即可表示为目标肝脏分割轮廓,进而计算出肝脏体积。结果评估结果显示,上述方法在肝脏体积误差方面表现出色,达到77分,但在局部（主要在肝脏尖端）出现较大的误差。结论该方法分割临床肝脏CT图像具有可行性。     

利用图谱匹配分割标注VHP数据集

利用TT脑图谱中丰富的结构信息,本文提出了一种自动分割脑图像的方法,并将其用于Visible Human数据集（VHD）的脑图像的分割,这种方法可分为两步,首先,将VHD中的脑图像和TT Atlas配准,通过图像和医学图谱的匹配,可以把图谱中存储的拓朴信息直接映射到ＶＨＤ,然后,利用这个预分割的模板对VHD脑图像进行模糊聚类分割,为自动将模板中的结构信息用于分割,本文利用Chamfer距离变换,提出了一中引入形状因子的FCM聚类算法。     

基于ITK多图谱配准的脑部皮质下核团的分割研究

大量研究表明,阿尔茨海默症(AD)的病变与大脑皮质下核团的萎缩息息相关,某些核团的萎缩(如海马)可能成为AD疾病早期诊断的标志,而皮质下核团的分割是研究核团萎缩模式的重要前提。基于AD患者和正常人各30例3DT1W-MR图像,先结合直方图分析和三维形态学分析方法对图像进行脑组织提取,后采用ITK配准算法将10个脑图谱图像经两阶段分别配准到提取脑组织后的图像空间。第一阶段实现基于均方差的仿射配准,第二阶段实现基于互信息的B样条形变配准,两阶段的配准均采用线性插值法和梯度下降的优化搜索方法。最后采用STAPLE融合算法,对配准后得到的10个目标图像进行图像融合,得到最终的分割结果。结果表明:除尾状核外,分割得到的其余6对核团的体积与常用的FSL-FIRST算法的分割结果无统计学差别(P>0.05);AD患者的右侧伏核和双侧海马发生萎缩(P<0.05)。因此,基于ITK配准框架的多图谱配准分割方法能有效分割MR图像上边界不明确的皮质下核团。     

结合脑图谱和水平集的MR图像分割的研究

本文利用脑图谱的先验知识并结合水平集等算法实现对脑MR图像的初步分割。主要步骤：（1）选取数字脑图谱，对图谱进行预处理；（2）实现图谱与脑MR图像的配准；（3）利用图谱提供的轮廓信息对水平集算法进行初始化，完成颅骨和脑脊液的提取以及脑白质和脑灰质的分割。实验结果表明，利用脑图谱提供的信息可有效解决水平集算法初始化问题，缩小求解空间，减少迭代次数，该方法具有较好的鲁棒性。     

医学图像的自动调窗与分割

计算机辅助外科系统中,图像引导的手术导航系统是一种技术含量最高、发展最快的外科手术设备,在神经外科、骨科、耳鼻喉科等有着广阔的应用前景。目前,手术导航系统在图像调窗、分割、配准均采用的是手工方式,迫切需要提高图像处理的自动化程度。本文提出的自动调窗与分割算法,比较好地实现了医学图像的自动调窗和自动分割功能:(1)通过对大量MRI和CT图像的灰度直方图分析,根据这一类图像的共性,给出相应的调窗算法,实验表明,该算法绝大部分MRI和CT图像的自动调窗效果接近于最佳;(2 )本文给出的种子生长分割方法,是基于灰度连通性原理,只需点击病灶,就可以将病灶及边缘准确地分割出来,并可以动态、实时地控制分割的效果,只要机器的内存允许,可以直接对三维图像进行三维分割。测试表明,该功能缩短和降低手术计划的时间和难度,提高导航手术的效率。     

一种基于纹理的颅脑CT图像病变检出算法

以颅脑CT图像为研究对象用基于纹理的数字化统计图谱方法进行了病变自动化检出的研究,提出并创建基于纹理的数字化统计图谱——纹理层析图谱。通过比较待诊断颅脑CT图像与纹理层析图谱间的差异,实现颅脑CT图像中多种病变的计算机自动化检出。实验结果表明,在不知道病变种类的前提下,基于纹理层析图谱的病变检出算法可以实现颅脑CT图像所含病变的自动化检出。利用图像的纹理信息变化是实现颅脑CT图像病变检出的一个有效途径。     

病变细胞显微图像分析与识别技术的研究

依据病变细胞的形态和颜色特征,我们提出了一种基于RGB和HIS彩色空间的自适应自动阈值分割算法,该算法能有效地将病变细胞的胞核从复杂的背景中提取出来.在分割图像的基础上,应用canny边缘检测算法提取出细胞边缘,采用八链码跟踪技术提取出细胞的特征值.为了同正常细胞比较,同时提取了正常细胞的特征值,并提出了二步识别算法以对正常和病变细胞进行识别.实验结果表明,该系统能有效地分割血细胞图像并且诊断率较高.     

基于VIP-Man图谱配准的腹部器官分割方法

提出了一种基于图谱配准的腹部器官分割方法.首先将一套预标记图谱向个体图像进行配准,建立二者之间器官的基本对应关系,同时完成对感兴趣器官的识别,其中配准包含全局配准和器官配准.然后,借助已配准的图谱,采用模糊连接方法对感兴趣器官进行分割.腹PCT和MR实验测试结果证明:这种方法实现了模糊连接分割方法中各项参数的自动指定,减轻了人工负担,提高了结果的可靠性.     

应用深度学习网络实现肾小球滤过膜超微病理图像的语义分割

肾小球滤过膜包含内皮细胞、肾小球基底膜和足细胞3层超微结构,其形态改变是诊断肾小球疾病的重要指标之一。准确的滤过膜语义分割有助于病理医生识别和判断滤过膜细微的病理改变,为相关的病理诊断提供帮助。由于肾小球滤过膜的超微病理图像不仅结构复杂而且灰度分辨率很低,传统的分割算法均只能对其中形态最简单的基底膜部分进行分割,分割精度也难以保证。本文提出基于深度学习网络DeepLab-v3的肾小球滤过膜自动语义分割算法,应用空洞卷积扩大感受野,控制图像的特征分辨率,再通过空洞空间金字塔池化来获得多尺度的图像信息,最终将肾小球滤过膜的3个组成部分同时分割出来,并均能达到较好的分割效果。本文通过对重要参数进行实验探究,使得平均分割准确度达到0.776,是目前效果相对较好的模型。     

带有预处理的计划CT和在线CT的形变配准方法

CT扫描中,水溶性碘造影的存在使得计划CT和在线CT图像中血管内的HU值出现非常大的偏差,从而导致计划CT和在线CT图像错配。针对该问题,本研究提出了一种基于预处理的计划CT和在线CT形变配准方法。首先,根据CT图像组织和结构的信息,利用阈值分割方法分割出血管,并将所有分割中最大的联通区域作为初始分割的强化血管;其次,利用分割得到的强化血管区域外扩5 mm,作为外扩的强化血管,并将血管用固定的HU值进行填充;最后,对完成填充后的图像利用Demons算法进行形变配准。实验结果显示本文提出的带有预处理的形变配准方法,可以较好地解决水溶性碘造影剂引起的CT错配问题。     

显微图象的快速拼接

作者提出了基于图象匹配的显微图象快速拼接的新方法。拼接分三步完成：１．生成利于配准的梯度图象。２．基于金字塔数据结构的快速匹配。３．进行边界处清除图象拼接处的阶梯。优点是：拼接由计算机自动完成，速度快、精度高，消除了人为误差，并且能提供人工无法实现的拼接处光滑连接。     

基于深度学习的2D/3D医学图像配准研究

2D/3D配准在临床诊断和手术导航规划中有着广泛的应用,可解决医学图像领域中不同维度图像存在信息缺失的问题,能辅助医生在术中精准定位患者的病灶。常规的2D/3D配准方法主要依赖于图像的灰度进行配准,但非常耗时,不利于临床实时性的需求,并且配准过程中容易陷入局部最优值。提出用深度学习的方法来解决2D/3D医学图像配准问题。采用一个基于深度学习的卷积神经网络,通过网络对数字影像重建技术(DRR)进行训练并自动学习图像特征,预测X光图像所对应的参数,从而实现配准。以人体骨盆的模型骨为实验对象,根据骨盆的CT数据生成36000张DRR图像作为训练集,同时通过C臂采集模型骨的50张X光图像作为验证。结果显示,深度学习算法在相关系数、归一化互信息、欧式距离3个精度评价指标上的测试值分别为0.82±0.07、0.32±0.03、61.56±10.91,而常规2D/3D算法对应的测试值分别为0.79±0.07、0.29±0.03、37.92±7.24,说明深度学习算法的配准精度优于常规2D/3D算法的配准精度,且不存在陷入局部最优值的问题。同时,深度学习的配准时间约为0.03s,远低于常规2D/3D配准的时间,可满足临床对于实时配准的需求,未来将进一步开展临床数据的2D/3D配准研究。     

Registration of electronic portal images for patient set-up verification

Images acquired from an electronic portal imaging device are aligned with digitally reconstructed radiographs (DRRs) or other portal images to verify patient positioning during radiation therapy. Most of the currently available computer aided registration methods are based on the manual placement of corresponding landmarks. The purpose of the paper is twofold: (a) the establishment of a methodology for patient set-up verification during radiotherapy based on the registration of electronic portal images, and (b) the evaluation of the proposed methodology in a clinical environment. The estimation of set-up errors, using the proposed methodology, can be accomplished by matching the portal image of the current fraction of the treatment with the portal image of the baseline treatment (reference portal image) using a nearly automated technique. The proposed registration method is tested on a number of phantom data as well as on data from four patients. The phantom data included portal images that corresponded to various positions of the phantom on the treatment couch. For each patient, a set of 30 portal images was used. For the phantom data (for both transverse and lateral portal images), the maximum absolute deviations of the translational shifts were within 1.5 mm, whereas the in-plane rotation angle error was less than 0.5 degrees. The two-way Anova revealed no statistical significant variability both within observer and between-observer measurements (P > 0.05). For the patient data, the mean values obtained with manual and the proposed registration methods were within 0.5 mm. In conclusion, the proposed registration method has been incorporated within a system, called ESTERR-PRO. Its image registration capability achieves high accuracy and both intra- and inter-user reproducibility. The system is fully operational within the Radiotherapy Department of 'HYGEIA' Hospital in Athens and it could be easily installed in any other clinical environment since it requires standardized hardware specifications and minimal human intervention.     

Medical image registration

Radiological images are increasingly being used in healthcare and medical research. There is, consequently, widespread interest in accurately relating information in the different images for diagnosis, treatment and basic science. This article reviews registration techniques used to solve this problem, and describes the wide variety of applications to which these techniques are applied. Applications of image registration include combining images of the same subject from different modalities, aligning temporal sequences of images to compensate for motion of the subject between scans, image guidance during interventions and aligning images from multiple subjects in cohort studies. Current registration algorithms can, in many cases, automatically register images that are related by a rigid body transformation (i.e. where tissue deformation can be ignored). There has also been substantial progress in non-rigid registration algorithms that can compensate for tissue deformation, or align images from different subjects. Nevertheless many registration problems remain unsolved, and this is likely to continue to be an active field of research in the future.     

Automated matching and segmentation of lymphoma on serial CT examinations

In patients with lymphoma, identification and quantification of the tumor extent on serial CT examinations is critical for assessing tumor response to therapy. In this paper, we present a computer method to automatically match and segment lymphomas in follow-up CT images. The method requires that target lymph nodes in baseline CT images be known. A fast, approximate alignment technique along the x, y, and axial directions is developed to provide a good initial condition for the subsequent fast free form deformation (FFD) registration of the baseline and the follow-up images. As a result of the registration, the deformed lymph node contours from the baseline images are used to automatically determine internal and external markers for the marker-controlled watershed segmentation performed in the follow-up images. We applied this automated registration and segmentation method retrospectively to 29 lymph nodes in 9 lymphoma patients treated in a clinical trial at our cancer center. A radiologist independently delineated all lymph nodes on all slices in the follow-up images and his manual contours served as the "gold standard" for evaluation of the method. Preliminary results showed that 26/29 (89.7%) lymph nodes were correctly matched; i.e., there was a geometrical overlap between the deformed lymph node from the baseline and its corresponding mass in the follow-up images. Of the matched 26 lymph nodes, 22 (84.6%) were successfully segmented; for these 22 lymph nodes, several metrics were calculated to quantify the method's performance. Among them, the average distance and the Hausdorff distance between the contours generated by the computer and those generated by the radiologist were 0.9 mm (stdev. 0.4 mm) and 3.9 mm (stdev. 2.1 mm), respectively.     

Expert systems and automatic diagnostic systems in histopathology--a review

In this decade, the pathological information system has gradually been settled in many hospitals in Japan. Pathological reports and images are now digitized and managed in the database, and are referred by clinicians at the peripherals. Tele-pathology is also developing; and its users are increasing. However, in many occasions, the problem solving in diagnostic pathology is completely dependent on the solo-pathologist. Considering the need for timely and efficient supports to the solo-pathologist, I reviewed the papers on the knowledge-based interactive expert systems. The interpretations of the histopathological images are dependent on the pathologist, and these expert systems have been evaluated as "educational". With the view of the success in the cytological screening, the development of "image-analysis-based" automatic "histopathological image" classifier has been on ongoing challenges. Our 3 years experience of the development of the pathological image classifier using the artificial neural networks technology is briefly presented. This classifier provides us a "fitting rate" for the individual diagnostic pattern of the breast tumors, such as "fibroadenoma pattern". The diagnosis assisting system with computer technology should provide pathologists, especially solo-pathologists, a useful tool for the quality assurance and improvement of pathological diagnosis.     

On the use of EPID-based implanted marker tracking for 4D radiotherapy

Four-dimensional (4D) radiotherapy delivery to dynamically moving tumors requires a real-time signal of the tumor position as a function of time so that the radiation beam can continuously track the tumor during the respiration cycle. The aim of this study was to develop and evaluate an electronic portal imaging device (EPID)-based marker-tracking system that can be used for real-time tumor targeting, or 4D radiotherapy. Three gold cylinders, 3 mm in length and 1 mm in diameter, were implanted in a dynamic lung phantom. The phantom range of motion was 4 cm with a 3-s "breathing" period. EPID image acquisition parameters were modified, allowing image acquisition in 0.1 s. Images of the stationary and moving phantom were acquired. Software was developed to segment automatically the marker positions from the EPID images. Images acquired in 0.1 s displayed higher noise and a lower signal-noise ratio than those obtained using regular (> 1 s) acquisition settings. However, the markers were still clearly visible on the 0.1-s images. The motion of the phantom blurred the images of the markers and further reduced the signal-noise ratio, though they could still be successfully segmented from the images in 10-30 ms of computation time. The positions of gold markers placed in the lung phantom were detected successfully, even for phantom velocities substantially higher than those observed for typical lung tumors. This study shows that using EPID-based marker tracking for 4D radiotherapy is feasible, however, changes in linear accelerator technology and EPID-based image acquisition as well as patient studies are required before this method can be implemented clinically.     

基于互信息与边缘梯度相结合的多模医学图像配准方法

针对传统互信息配准方法未利用图像空间信息的缺点,本文研究了图像边缘信息的梯度相似性.首先采用小波模极大值边缘检测提取出图像边缘,提出将边缘图像的梯度相似性系数与传统的互信息相乘作为图像配准的目标函数.然后通过使用Powell优化算法对目标函数进行寻优,得出配准变换参数.最后在互信息的基础上引入图像边缘梯度信息,突出了全局最优解.实验结果表明,该方法可以得到精确、有效的配准结果.