基于改进Demons算法的非刚性配准研究及其在肿瘤放疗中的初步应用

提出一种改进的Demons非刚性配准算法,验证算法的有效性,并将其应用于图像引导放射治疗（image-guidedradiotherapy, IGRT）中治疗图像和计划图像的配准。方法 基于Brox等提出的梯度恒定假设和Malis的高效二阶最小化算法的思想,将图像灰度梯度场的相似性加入原始的能量函数中,推导出更新变形场的公式。利用有限内存的BFGS算法优化能量函数,自动确定迭代次数。分别利用模拟形变图像、变形体模图像和肝癌病人的临床CT图像验证改进算法的配准精度。结果 改进的Demons变形配准算法与原始的“Additive Demons”算法相比,配准精度更高,收敛速度更快。结论 在放射治疗的不同分次扫描过程中,因实际扫描条件的影响,两幅待配准图像像素灰度值范围不同的情况下,改进的Demons算法能够更好地实现快速精确的配准。     

基于改进Demons算法的非刚性配准及其在肿瘤放疗中的应用

目的提出一种改进的Demons非刚性配准算法,验证算法的有效性,并将其应用于图像引导放射治疗(image-guidedradiotherapy,IGRT)中治疗图像和计划图像的配准。方法基于Brox等提出的梯度恒定假设和Malis的高效二阶最小化算法的思想,将图像灰度梯度场的相似性加入原始的能量函数中,推导出更新变形场的公式。利用有限内存的BFGS算法优化能量函数,自动确定迭代次数。分别利用模拟形变图像、变形体模图像和肝癌病人的临床CT图像验证改进算法的配准精度。结果改进的Demons变形配准算法与原始的"Additive Demons"算法相比,配准精度更高,收敛速度更快。结论在放射治疗的不同分次扫描过程中,因实际扫描条件的影响,两幅待配准图像像素灰度值范围不同的情况下,改进的Demons算法能够更好地实现快速精确的配准。     

改进Demons算法的验证及其在4D-CT轮廓线推衍中的应用

目的 研究一种改进的Demons变形配准算法,验证算法的有效性,评价配准算法的精度,并将其应用于4D-CT轮廓线的推衍.方法 为加快收敛的速度和提高配准精度,将Demons算法的双向作用力进行重分配,并提出一个能量函数作为相似性测度,利用BFGS优化算法对能量函数进行优化,避免了经典算法中需指定迭代次数的缺陷.分别利用模拟形变CT图像与变形模体,分析和测量了配准算法的精度,并对该算法在4D-CT轮廓线推衍中的应用进行了评价.结果 改进后的Demons算法与经典的Demons算法以及类似的变形配准算法相比,有较高的配准精度,推衍生成的轮廓线与人工勾画的轮廓线匹配程度高.结论 变形配准是4D-CT中的一项关键技术,改进后的Demons算法应用于轮廓线的推衍有望大大治疗计划设计的轮廓勾画工作量,其配准精度能满足实际临床的需要.     

基于改进Demons算法的非刚性配准及其在肿瘤放疗中的应用

摘要：目的提出一种改进的Demons 非刚性配准算法,验证算法的有效性,并将其应用于图像引导放射治疗（image-guided
radiotherapy, IGRT）中治疗图像和计划图像的配准。方法基于Brox等提出的梯度恒定假设和Malis的高效二阶最小化算法的
思想,将图像灰度梯度场的相似性加入原始的能量函数中,推导出更新变形场的公式。利用有限内存的BFGS算法优化能量函
数,自动确定迭代次数。分别利用模拟形变图像、变形体模图像和肝癌病人的临床CT图像验证改进算法的配准精度。结果改
进的Demons变形配准算法与原始的“Additive Demons”算法相比,配准精度更高,收敛速度更快。结论在放射治疗的不同分
次扫描过程中,因实际扫描条件的影响,两幅待配准图像像素灰度值范围不同的情况下,改进的Demons算法能够更好地实现快
速精确的配准。
     

基于Demons算法的MR图像与病理切片的非刚性配准

目的:采用Demons算法实现脑胶质瘤MR图像与病理切片图像的非刚性配准。方法:采用Demons算法,加入局部结构张量信息,构建能量函数以获得新的形变向量,最后通过迭代实现MR图像与病理切片图像的配准。结果:配准后病理图像的肿瘤轮廓与MR图像基本吻合。结论:Demons算法可实现脑胶质瘤MR图像与病理切片的非刚性配准,配准后两幅图像的肿瘤大小信息基本达到一致,可为进一步的模型构建提供参考。     

虚拟中国人女性一号图像数据的配准

目的 解决中国数字化女性虚拟人图像数据三维重建前的配准问题。方法 根据图像数据特点，应用基于外置标记点和基于轮廓比较的最小二乘法两种方法进行参数计算，依参数对图像进行刚体变换处理完成配准。结果 使用这两种方法对数据集图像进行配准处理，达到了预期效果。结论 这两种方法能够实现此数据集图像的精确配准．而且具有运算量小且易于编程实现的特点。     

虚拟中国人女性一号宫颈的任意剖面显示的研究

目的:研究中国数字化女性虚拟人宫颈的剖面显示问题.方法:根据图像数据特点,从所获得的女性骨盆数据中提取出女性宫颈数据,然后通过计算机相邻两层图像之间像素对的相同程度来对其进行平移配准,最后利用C Builder语言编程实现剖面显示.结果:使用文中提及的方法对图像进行配准处理,我们发现在配准搜索过程中出现局部相似性最大的情况,从而使得搜索提前结束.结论:我们建议采用基于最大互信息量配准的方法,通过统计方法来表达图像像素的相似性特征,并利用优化算法有效克服局部极值,以便提高图像显示的精度,从而较好地应用于虚拟人体的其他部位的剖面显示.     

虚拟中国人女性一号图像数据的配准

目的解决中国数字化女性虚拟人图像数据三维重建前的配准问题。方法根据图像数据特点,应用基于外置标记点和基于轮廓比较的最小二乘法两种方法进行参数计算,依参数对图像进行刚体变换处理完成配准。结果使用这两种方法对数据集图像进行配准处理,达到了预期效果。结论这两种方法能够实现此数据集图像的精确配准,而且具有运算量小且易于编程实现的特点。     

基于数字化可视人数据集的医学图像配准研究

目的:采用医学图像配准技术,对中国数字化可视人数据集进行准确、高效的配准。方法:①图像Lab色度空间变换和二值化;②图像分割和特征量提取;③构造坐标变换矩阵;④图像序列配准变换。结果:配准后的图像序列中,人体解剖结构在空间和结构上实现了精确的匹配对应。进行三维重建后,几何模型外表面具有较高的光滑度,说明方法有效、可靠。结论:采用特征提取和快速坐标变换的配准方法可较好地应用于数字化可视人数据集的图像配准。     

基于引导滤波的多图谱医学图像分割

目的为了有效的利用图谱的先验信息和待分割图像的灰度信息,并在融合标号图像的过程中校正配准引起的误差,得到
光滑、准确的分割结果,提出了一种新的基于引导滤波的多图谱医学图像分割方法。方法本文将多图谱配准与引导滤波相结
合。该方法包含4个部分：第一部分为多图谱配准,通过配准将图谱中存储的形状先验信息映射到待分割图像;第二部为标号融
合,利用配准的相似性作为权重,将形变后的标号图像融合在一起;第三部分为引导滤波,利用引导滤波引入待分割图像的灰度
信息,可以校正配准引起的误差;最后通过阈值处理,得到最终的分割结果。结果对15例脑部MR图像数据中的海马体进行分
割实验,左、右海马体分别达到了86%及87.4%的分割精度,与传统的标号融合算法相比,平均分割精度提升了2.4%。结论本
文方法结合多配谱配准与引导滤波的优势,提高了海马的分割精度,并得到光滑有效的分割精度。
     

Content-based image retrieval for Lung Nodule Classification Using Texture Features and Learned Distance Metric

Similarity measurement of lung nodules is a critical component in content-based image retrieval (CBIR), which can be useful in differentiating between benign and malignant lung nodules on computer tomography (CT). This paper proposes a new two-step CBIR scheme (TSCBIR) for computer-aided diagnosis of lung nodules. Two similarity metrics, semantic relevance and visual similarity, are introduced to measure the similarity of different nodules. The first step is to search for K most similar reference ROIs for each queried ROI with the semantic relevance metric. The second step is to weight each retrieved ROI based on its visual similarity to the queried ROI. The probability is computed to predict the likelihood of the queried ROI depicting a malignant lesion. In order to verify the feasibility of the proposed algorithm, a lung nodule dataset including 366 nodule regions of interest (ROIs) is assembled from LIDC-IDRI lung images on CT scans. Three groups of texture features are implemented to represent a nodule ROI. Our experimental results on the assembled lung nodule dataset show good performance improvement over existing popular classifiers.     

基于结构光扫描的多区域配准技术用于脊柱外科三维手术导航的试验研究

目的 建立基于结构光扫描的脊柱手术导航配准系统.方法 获得胸椎模型的CT扫描图像及结构光扫描图像后,结合预配准与多区域迭代最近点(ICP)算法的精细配准方法,对CT图像及结构光图像进行配准.选择胸椎上4个区域进行不同组合,计算配准精度差异.对结构光扫描施加位移及三维正态噪声,观察配准精度变化.通过小牛脊柱标本进一步验证配准精度.结果 结合预配准与多区域配准的方法可实现脊椎骨可见光数据与CT数据的配准,选用两个及两个以上区域进行组合配准的误差＜1 mm,结构光扫描受到干扰后,配准误差仍＜1 mm.结论 基于结构光的预配准结合多区域配准算法可实现结构光数据与CT数据的配准,有着高精度、抗干扰能力强的优点,较传统的配准方法更为精确.

Abstract：

Objective To study the registration method based on structured light scanning for navigation assisted spinal surgery and assess its accuracy so as to construct a registration system for the navigation assisted spinal surgery using structured light scanning. Methods Both the computed tomographic (CT) dataset and the structured light scanning images of thoracic vertebra were obtained. The preregistration and multi-segment iterative closest point (ICP) algorithm were used for the registration of CT images and structured light images. Four segmentations were selected from the surface of thoracic vertebra and placed into different combinations. The accuracy for each combination was studied. Noise and perturbation were exerted to structured light and registration accuracy was studied. And calf vertebra was used for further verification. Results A combination of pre-registration and multi-segment iterative closest point (ICP) algorithm was competent for the registration of CT scanning data and the structured light scanning data. The registration error was less than 1 mm when two and more segments were selected for registration combination. The registration error was less than 1 mm when noise was exerted. Conclusion With a high accuracy and a perturbation resistance, a combination of pre-registration and multi-segment registration algorithm based on structured light scanning is competent for the registration of CT scanning data and structured light scanning data.     

基于术前CT影像组学列线图可预测Ⅰ～Ⅲ期肾透明性细胞癌术后复发

目的 探讨Ⅰ~Ⅲ期肾透明细胞癌术后复发的术前CT影像组学特征并构建列线图,以期为肾癌个体化治疗提供参考。方法 回顾性收集256例（训练集175例,测试集81例）肾透明细胞癌患者的临床病理及 CT 资料。利用 ITK-SNAP 软件和PyRadiomics计算平台对肿瘤的容积图像进行分割和特征提取。训练集中,基于lasso-CV算法进行特征筛选,并计算影像组学评分Rad_score;利用单因素和多因素逻辑回归分析筛选临床病理及CT特征为Clinic因素;构建Rad_score、Clinic、Rad_score+Clinic列线图,并在测试集中进行验证。评估列线图的辨别度和校准度,应用决策曲线分析评估其临床应用价值。结果 6个影像组学特征最终用于计算Rad_score。Clinic因素为KPS评分、血小板、钙化和TNM临床分期。在辨别度方面,Rad_score+Clinic列线图的效能（训练集AUC 0.84,测试集AUC 0.85）显著高于Rad_score列线图（训练集AUC 0.78,P=0.029;测试集AUC0.77,P=0.025）和 Clinic列线图（训练集AUC 0.77,P=0.014,测试集AUC 0.77,P=0.011）。校准度方面,Rad_score+Clinic列线图拟合优度检验为训练集P=0.065,测试集P=0.628。决策曲线分析显示,加入Rad_score后的Rad_score+Clinic列线图比单纯Clinic列线图应用价值高。结论 基于术前CT影像组学特征的列线图预测Ⅰ~Ⅲ期肾透明细胞癌术后复发有较高的效能,可为肾癌个体化治疗提供参考。     

Application of Attribute Weighting Method Based on Clustering Centers to Discrimination of Linearly Non-Separable Medical Datasets

In this paper, attribute weighting method based on the cluster centers with aim of increasing the discrimination between classes has been proposed and applied to nonlinear separable datasets including two medical datasets (mammographic mass dataset and bupa liver disorders dataset) and 2-D spiral dataset. The goals of this method are to gather the data points near to cluster center all together to transform from nonlinear separable datasets to linear separable dataset. As clustering algorithm, k-means clustering, fuzzy c-means clustering, and subtractive clustering have been used. The proposed attribute weighting methods are k-means clustering based attribute weighting (KMCBAW), fuzzy c-means clustering based attribute weighting (FCMCBAW), and subtractive clustering based attribute weighting (SCBAW) and used prior to classifier algorithms including C4.5 decision tree and adaptive neuro-fuzzy inference system (ANFIS). To evaluate the proposed method, the recall, precision value, true negative rate (TNR), G-mean1, G-mean2, f-measure, and classification accuracy have been used. The results have shown that the best attribute weighting method was the subtractive clustering based attribute weighting with respect to classification performance in the classification of three used datasets.     

基于双能X射线透视成像的肺部肿瘤运动跟踪方法及临床评价

目的:利用双能X射线透视成像技术,通过呼吸周期内高低能运动图像序列对软组织剪影图像配准,提供一种无需金属标记的肺部肿瘤呼吸运动跟踪方法。方法:以肺部肿瘤患者的高低能X射线透视图像为研究对象,通过自动双能减影算法获得软组织减影图像,然后采用本研究提出的一种结合自适应参考图像选择和归一化互信息匹配的肿瘤运动跟踪算法,计算肿瘤呼吸运动轨迹和运动幅度。采集并分析19例肺癌患者的临床数据,以人工测量结果为参考基准,评价肿瘤运动跟踪算法的准确性。结果:19例病例分析结果显示,运动跟踪算法计算获得的肿瘤呼吸运动轨迹和运动幅度,与人工测量方法获得的结果具有很好的一致性。对大部分病例,头脚方向运动幅度明显大于左右和腹背方向运动幅度,位于肺下半部分肿瘤的运动幅度明显大于位于肺中上部肿瘤。结论:无需金属标记的肿瘤运动跟踪算法,利用双能减影软组织图像,直接对肿瘤进行图像配准,能够较准确地跟踪肺部肿瘤呼吸运动。     

Generative Adversarial Network for Medical Images (MI-GAN)

Deep learning algorithms produces state-of-the-art results for different machine learning and computer vision tasks. To perform well on a given task, these algorithms require large dataset for training. However, deep learning algorithms lack generalization and suffer from over-fitting whenever trained on small dataset, especially when one is dealing with medical images. For supervised image analysis in medical imaging, having image data along with their corresponding annotated ground-truths is costly as well as time consuming since annotations of the data is done by medical experts manually. In this paper, we propose a new Generative Adversarial Network for Medical Imaging (MI-GAN). The MI-GAN generates synthetic medical images and their segmented masks, which can then be used for the application of supervised analysis of medical images. Particularly, we present MI-GAN for synthesis of retinal images. The proposed method generates precise segmented images better than the existing techniques. The proposed model achieves a dice coefficient of 0.837 on STARE dataset and 0.832 on DRIVE dataset which is state-of-the-art performance on both the datasets.     

椭球先验约束的前列腺磁共振图像分割

目的 为了有效的利用图谱的先验信息和待分割图像的灰度信息,提出一种新的椭球先验约束下的前列腺MR图像多图谱分割算法.方法 将多图谱分割与椭球形状先验相结合,在多图谱分割过程中引入椭球先验知识,针对椭球先验约束下的前列腺感兴趣区域进行图谱选择,大大避免了前列腺周围组织与器官对图谱选择造成的干扰;其次,在图谱融合过程中加入椭球先验项进行约束,对通过配准技术引入的前列腺图谱形状先验进行校正和补偿,有效避免了由配准误差引起的错误分割的情况.结果 对50例前列腺MR图像进行分割实验,实验结果表明该算法对前列腺数据的分割精度均在80％以上,平均精度提高到了88.12％.结论 椭球先验约束的前列腺MR图像多图谱分割算法稳定有效,分割结果精确度高.     

Active learning for clinical text classification: is it better than random sampling?

Objective

This study explores active learning algorithms as a way to reduce the requirements for large training sets in medical text classification tasks.

Design

Three existing active learning algorithms (distance-based (DIST), diversity-based (DIV), and a combination of both (CMB)) were used to classify text from five datasets. The performance of these algorithms was compared to that of passive learning on the five datasets. We then conducted a novel investigation of the interaction between dataset characteristics and the performance results.

Measurements

Classification accuracy and area under receiver operating characteristics (ROC) curves for each algorithm at different sample sizes were generated. The performance of active learning algorithms was compared with that of passive learning using a weighted mean of paired differences. To determine why the performance varies on different datasets, we measured the diversity and uncertainty of each dataset using relative entropy and correlated the results with the performance differences.

Results

The DIST and CMB algorithms performed better than passive learning. With a statistical significance level set at 0.05, DIST outperformed passive learning in all five datasets, while CMB was found to be better than passive learning in four datasets. We found strong correlations between the dataset diversity and the DIV performance, as well as the dataset uncertainty and the performance of the DIST algorithm.

Conclusion

For medical text classification, appropriate active learning algorithms can yield performance comparable to that of passive learning with considerably smaller training sets. In particular, our results suggest that DIV performs better on data with higher diversity and DIST on data with lower uncertainty.     

基于自适应Unet网络的鼻咽癌放疗危及器官自动分割方法

目的 探讨鼻咽癌放射治疗中的危及器官（OARs）的自动分割的准确性。方法 在自动分割模型研究中,经CT扫描和医生手动分割后,选取147例鼻咽癌患者的CT图像及其对应勾画的OARs结构,并对其进行完全随机化分组,分成训练集（115例）、验证集（12例）、测试集（20例）。采用自适应直方图均衡化对CT图像进行预处理。利用端到端训练提高建模效率,实现一种基于三维Unet的改进网络（AUnet）,将器官大小作为先验知识引入卷积核大小设计中,使网络能自适应地提取不同大小器官的特征,从而提高模型的性能。比较自动与手动分割的DSC（Dice Similarity Coefficient）系数和豪斯多夫（HD）距离以验证AUnet网络的有效性。结果 测试集的平均DSC和HD分别为0.86±0.02和4.0±2.0 mm。除视神经、视交叉外,AUnet与手动分割结果无统计学差异（P>0.05）。结论 引入自适应机制后,AUnet能较为准确地实现基于CT图像对鼻咽癌的危及器官的自动分割,临床应用中可大幅度提高医生的工作效率及分割的一致性。