肝硬化的超声特征分析

利用肝脏硬度的无创估计系统，可以及时计肝脏的弹性状况，对了解和控制肝病进程有积极的意义，本文先在常用超声仪上获得肝脏的M型超声图像，接着从图像中分别提取肝脏包膜的运动包络和A型扫描信号相关系数随时间变化的曲线，然后从运动包络和相关曲线中提取出相应的特征参量，最后用Fisher判别法进行肝脏是否硬化的分类决策，结果表明：该方法可以为简单，有效地诊断肝硬化提供新的手段。     

基于形态特征判别超声图像中乳腺肿瘤的良恶性

乳腺肿瘤超声图像的形态特征对判别肿瘤的良恶性具有重要的价值。为提高乳腺肿瘤超声诊断的准确率,提出一种基于其形态特征进行分类判别的计算机辅助诊断系统。该系统首先采用灰度阈值分割和动态规划相结合的方法提取超声图像中乳腺肿瘤的边缘,然后对所得边缘计算相应的三种形态参数,最后分别采用Fisher线性判据、误差反向传播神经网络和径向基函数神经网络对形态参数进行分类。该系统在157幅乳腺肿瘤(包括良性81例、恶性76例)超声图像上训练和测试,三种分类器均能取得较高的判别精度,其中误差反向传播神经网络和径向基函数神经网络的判别准确率、敏感性和特异性分别高达94.95 %、95.74%和94.23%。结果表明,基于乳腺肿瘤超声图像的形态特征建立的神经网络系统对肿瘤的良恶性具有较好的判别能力。     

一种超声乳腺图像中提取肿瘤边缘的方法

乳腺肿瘤边缘的提取是基于肿瘤超声图像特征进行良恶性诊断的基础.为此提出一种基于小波变换和动态规划从超声乳腺图像中提取肿瘤边缘的方法.先采用小波变换法提取乳腺肿瘤超声图像的初始边缘,再结合图像梯度的动态规划法对初始边缘进行修正,提取更为准确的乳腺肿瘤边缘.通过对临床采集的30例超声乳腺图像肿瘤边缘的提取和分析研究,表明该方法可以有效地用于超声乳腺图像肿瘤边缘的提取.     

超光谱遥感图像特征分析

超光谱遥感图像是一种三维立体图像,可反映地物的空间信息和光谱信息,其数据量庞大.本文通过对实际多光谱遥感图像的测试,分析了空间相关性和谱间相关性;计算出自相关系数和互相关系数;并与普通图像进行比较,揭示了超光谱图像的特征,为超光谱图像处理和压缩编码奠定了理论基础.     

改进型脉冲耦合神经网络检测乳腺肿瘤超声图像感兴趣区域

为了解决超声图像斑点噪声、伪影、低图像对比度和图像亮度不均匀等问题,提出了一种改进的简化脉冲耦合神经网络(SPCNN)结合模糊互信息量的方法来自动检测乳腺肿瘤超声图像的感兴趣区域(ROI).首先,对超声图像进行模糊增强预处理;然后,通过改进SPCNN对超声图像进行点火,以最大模糊互信息量作为最优判决准则,获得相应的分类...     

颈部淋巴结超声图像的计算机辅助分析

淋巴结的超声检查对疾病的诊断有重要意义,本文对颈部淋巴结超声图像的计算机辅助分析进行研究.首先根据颈部淋巴结的医学超声特性,对每一个淋巴结提取9个量化特征参数,然后基于K近邻法将其判为良性或恶性.采用留一法对62个病例进行测试,当k=5时,识别正确率为88.71%,灵敏度为83.87%,特异度为93.55%.本文为颈部淋巴结疾病的超声诊断提供一种客观、简捷的手段.     

超声信号的特征提取与选择在缺陷分类识别中的应用

超声缺陷回波信号的特征提取与选择是超声波检测的基础和关键。结合待检测超声信号的特点,在时域、频域上研究了超声缺陷信号的特征提取,提出SBS和SFS结合的特征选择方法。在满足识别准确率的同时,有效地降低了分类器输入向量的维数,提高了运算效率。     

基于血管内超声图像自动识别易损斑块

为克服手工判别动脉粥样硬化易损斑块耗时耗力、主观性强、重复性差等缺点,研究了基于血管内超声自动识别易损斑块的方法.首先将Contourlet变换与Snake模型相结合进行斑块的图像分割,提取内腔轮廓与外弹力膜.接着实现经典形态特征的计算机自动提取,并提取纹理、弹性两类新特征以量化斑块属性,其中纹理特征包括一阶统计量和灰...     

基于迁移学习与特征融合的眼底图像分类

针对大量眼底图片难以收集和标注、有经验的眼科医生地区分配不均匀等,导致眼底疾病患者检查准确度低、花费时间较长等问题,本文基于迁移学习提出一种图像分类方法:首先修改EfficientNet-B0和EfficientNet-B7模型并进行参数微调,将微调后的模型作为特征提取器用于提取眼底图像的特征,再对提取的特征进行特征融...     

原发性输卵管癌的超声图像特征及误诊分析

目的:观察原发性输卵管癌(PFTC)的超声图像特征,分析误诊原因,总结临床诊断思路。方法 :于我院接受外科手术,经术后病理组织学检查明确PFTC诊断的患者53例,调取患者术前超声检查资料,对超声分型表现及诊断结果进行分析,并归纳PFTC超声误诊率及误诊原因。结果:53例PFTC患者中,超声分型Ⅰ型8例,Ⅱ型3例,Ⅲ型17例,Ⅳ型18例,Ⅴ型7例。超声分型Ⅰ型、Ⅱ型者术前超声准确诊断率均为100.00%;Ⅲ型诊断准确率82.35%,3例患者误诊为浆膜下肌瘤;Ⅳ型患者转移率为100.00%,均误诊为卵巢来源恶性肿瘤,误诊率100%;Ⅴ型术前超声均漏诊。结论:腊肠状囊性肿块是PFTC的典型征象,对于不典型图像根据输卵管病变位置、大小、内部回声及周围脏器解剖关系及病史综合判断,有望降低误诊率、漏诊率。     

超声乳腺肿瘤图像中种子点的自动定位研究

利用种子区域增长对超声乳腺肿瘤图像进行分割是一种常用的计算机辅助诊断方法。为实现种子点的自动快速定位,满足实时在线分割图像的需求,根据超声乳腺肿瘤图像的结构特征,综合图像的灰度因素和空间因素,提出了一种基于迭代四叉树分解的算法。该算法将满足特定阈值的图像分裂转化为寻找种子区域,以实现种子点的自动定位。对105幅超声乳腺肿瘤图像进行了实验验证,结果表明,该算法准确率能够达到94.28%,平均耗时2.97 s,不但满足了种子点的自动定位于图像肿瘤内部,而且需要调整的参数少,其定位效率要高于人工选择。     

Optimizing the transfer-learning with pretrained deep convolutional neural networks for first stage breast tumor diagnosis using breast ultrasound visual images

Female accounts for approximately 50% of the total population worldwide and many of them had breast cancer. Computer-aided diagnosis frameworks could reduce the number of needless biopsies and the workload of radiologists. This research aims to detect benign and malignant tumors automatically using breast ultrasound (BUS) images. Accordingly, two pretrained deep convolutional neural network (CNN) models were employed for transfer learning using BUS images like AlexNet and DenseNet201. A total of 697 BUS images containing benign and malignant tumors are preprocessed and performed classification tasks using the transfer learning-based CNN models. The classification accuracy of the benign and malignant tasks is completed and achieved 92.8% accuracy using the DensNet201 model. The results thus achieved compared in state of the art using benchmark data set and concluded proposed model outperforms in accuracy from first stage breast tumor diagnosis. Finally, the proposed model could help radiologists diagnose benign and malignant tumors swiftly by screening suspected patients.     

基于相位和GGVF的水平集乳腺超声图像分割

提出一种基于相位和广义梯度矢量流(generalized gradient vector flow,GGVF)的水平集分割方法,并用于乳腺超声图像的肿瘤分割。首先,在频域空间,用Cauchy核替代Log-Gabor作为正交滤波器对图像进行滤波,提取来自于单演信号的多尺度图像特征,引入相位一致的思想将多尺度特征结合起来进行边界检测;然后,在此基础上,利用相位一致梯度图定义了一个基于相位的速度停止项函数,同时改进了GGVF;最后,将得到的速度停止项和梯度矢量流融入到水平集演化方程中来控制曲线的演化,获得乳腺肿瘤的边界。实验结果表明,使用该分割方法可获得比现有方法更好的乳腺肿瘤分割结果。     

Speckle reduction in breast cancer ultrasound images by using homogeneity modified bayes shrink

Ultrasound imaging suffers from severe artifacts caused by speckle noise. The paper introduces an algorithm for speckle noise reduction in breast cancer ultrasound images. Based on wavelet analysis and filtering, we employed a combination of homogeneity filtering and modified bayes shrink methods to remove noise while keeping the sharpness of important features. First, we replace pixel intensity by the mean of homogenous neighborhood and then, the threshold value of modified bayes shrink is employed to distinguish homogenous regions from regions with speckle noise obtained from homogeneity filtering. The proposed algorithm is called Homogeneity Modified Bayes Shrink (HMBS). A comparative study with other despeckling methods, using quantitative indices, showed the superiority of the proposed method over those methods.     

Localization of spots in FISH images of breast cancer using 3‐D shape analysis

The fluorescence in situ (FISH) belongs to the most often used molecular cytogenetic techniques, applied in many areas of diagnosis and research. The analysis of FISH images relies on localization and counting the red and green spots in order to determine HER2 status of the breast cancer samples. The algorithm of spot localization presented in the paper is based on 3‐D shape analysis of the image objects. The subsequent regions of the image are matched to the reference pattern and the results of this matching influence localization of spots. The paper compares different shapes of the reference pattern and their efficiency in spot localization. The numerical experiments have been performed on the basis of 12 cases (patients), each represented by three images. Few thousands of cells have been analysed. The quantitative analyses comparing different versions of algorithm are presented and compared to the expert results. The best version of the procedure provides the absolute relative difference to the expert results smaller than 3%. These results confirm high efficiency of the proposed approach to the spot identification. The proposed method of FISH image analysis improves the efficiency of detecting fluorescent signals in FISH images. The evaluation results are encouraging for further testing of the developed automatic system directed to application in medical practice.