改进标记分水岭的乳腺X线图像肿块检测方法

乳腺X线图像肿块大小不一,固定参数的传统标记分水岭算法无法实现乳腺X线图像肿块的有效检测。针对此问题,文中提出了一种结合形状特征和改进型标记分水岭的乳腺X线图像肿块检测方法。在计算前景标记时,结合标记的形状特征判定前景标记,通过对前景标记进行形态学膨胀并提取边缘以获得背景标记,利用改进型自适应参数标记分水岭算法实现肿块检测。实验结果表明,文中算法通过结合形状特征,自适应地选择合适的形态学参数,使得肿块检测准确率高于传统标记分水岭算法。     

SVM算法及其在乳腺X片微钙化点 自动检测中的应用

支持矢量机(SVM)是一种新的统计学习方法,其学习原则是使结构风险最小,而非经典学习方法所遵循经验风险最小原则.这使得SVM具有更强的泛化能力.并且,由于SVM求解的是凸二次优化问题,使之能保证所找到的极值解就是全局最优解.本文首次将SVM算法用于乳腺X影像微钙化点自动检测中,对临床实际病例的试用结果表明,同目前常用的基于经验风险最小的人工神经网络(ANN)分类方法相比,SVM具有更高的识别率,值得应用推广.     

Mammogram Enhancement Using Lifting Dyadic Wavelet Transform and Normalized Tsallis Entropy

In this paper, we present a new technique for mammogram enhancement using fast dyadic wavelet transform （FDyWT） based on lifted spline dyadic wavelets and normalized Tsallis entropy. First, a mammogram image is decom- posed into a multiscale hierarchy of low-subband and high-subband images using FDyWT. Then noise is suppressed using normalized Tsallis entropy of the local variance of the modulus of oriented high-subband images. After that, the wavelet coefficients of high-subbands are modified using a non-linear operator and finally the low-subband image at the first scale is modified with power law transformation to suppress background. Though FDyWT is shift-invariant and has better poten- tial for detecting singularities like edges, its performance depends on the choice of dyadic wavclcts. On the other hand, the nulnber of vanishing moments is an important characteristic of dyadic wavelets for singularity analysis because it provides an upper bound measurement for singularity characterization. Using lifting dyadic schemes, we construct lifted spline dyadic wavelets of different degrees with increased number of vanishing moments. We also examine the effect of these wavelets on mammogram enhancement. The method is tested on mammogram images, taken from MIAS （Mammographic Image Analysis Society） database, having various background tissue types and containing different abnormalities. The comparison with tile state-of-the-art contrast enhancement methods reveals that the proposed method performs better and the difference is statistically significant.     

基于注意力机制与迁移学习的乳腺钼靶肿块分类

针对乳腺钼靶图像中良恶性肿块难以诊断的问题,提出一种基于注意力机制与迁移学习的乳腺钼靶肿块分类方法,并用于医学影像中乳腺钼靶肿块的良恶性分类.首先,构建一种新的网络模型,该模型将注意力机制CBAM(Convolutional Block Attention Module)与残差网络ResNet50相结合,用于提高网络对...     

一种新颖的乳腺X线影像中钙化点检测方法

乳腺癌是妇女常见恶性肿瘤之一,早期诊断和早期治疗是降低乳腺癌患者死亡率的关键。微钙化是乳腺癌早期的一个重要标志,因此,快速准确地找出乳腺X光片中的钙化点成为成功诊断的第一步。现有多种方法能用于检测钙化点并各有优缺点,其中典型的高斯-拉普拉斯算子（LOG）是有效方法之一,尽管其能较精确地检出钙化点的位置但检测效率低。级联形态学滤波算子的LOG改进了LOG的效率,但仍无法满足大规模普查的高效率要求。通过提出一维和二维LOG相级联的方法来实现高效实时的钙化点的检测,并通过实验证实了所提检测方法的有效性。     

基于深度语义模型的乳腺X线图像检索

图像特征是基于内容的图像检索（Content-based image retrieval,CBIR）的关键,大部分使用的手工特征难以有效地表示乳腺肿块的特征,底层特征与高层语义之间存在语义鸿沟。为了提高CBIR的检索性能,本文采用深度学习来提取图像的高层语义特征。由于乳腺X线图像的深度卷积特征在空间和特征维度上存在一定的冗余和噪声,本文在词汇树和倒排文件的基础上,对深度特征的空间和语义进行优化,构建了两种不同的深度语义树。为了充分发挥深度卷积特征的识别能力,根据乳腺图像深度特征的局部特性对树节点的权重进行细化,提出了两种节点加权方法,得到了更好的检索结果。本文从乳腺X线图像数据库（Digital database for screening mammography, DDSM）中提取了2 200个感兴趣区域（Region of interest,ROIs）作为数据集,实验结果表明,该方法能够有效提高感兴趣肿块区域的检索精度和分类准确率,并且具有良好的可扩展性。     

Detection and classification of microcalcification from digital mammograms with firefly algorithm,extreme learning machine and non-linear regression models: A comparison

In this study, abnormalities in medical images are analysed using three classifiers, and the results are compared. Breast cancer remains a major public health problem among women worldwide. Recently, many algorithms have evolved for the investigation of breast cancer diagnosis through medical imaging. A computer-aided microcalcification detection method is proposed to categorise the nature of breast cancer as either benign or malignant from input mammogram images. The standard mammogram image corpus, the Mammogram Image Analysis Society database is utilised, and feature extraction is performed using five different wavelet families at level 4 and level 6 decomposition. The work is accomplished through firefly algorithm (FA), extreme learning machine (ELM) and least-square-based non-linear regression (LSNLR) classifiers. The performance of the classifiers is compared by benchmark metrics, such as total error rate, specificity, sensitivity, area under the receiver operating characteristic curve, precision, F1 score and the Matthews correlation coefficient. As validation of the classifier results, a kappa analysis is included to determine the agreement among classifiers. The LSNLR classifier attains a 3% to 7% improvement in average accuracy compared with the average classification accuracy of the FA (86.75%) and ELM (90.836%) classifiers.     

基于迁移学习的多视角乳腺肿块和钙化簇分类方法

针对乳腺肿块和钙化簇分类任务中可用训练数据量较少的问题,结合乳腺钼靶图成像特点提出了一种基于二次迁移学习的多视角模型。首先,使用CBIS-DDSM制作乳腺局部组织切片数据集来预训练主干网络,完成主干网络的领域适应性学习,使之具备基本的病理特征捕捉能力;随后,把主干网络二次迁移到多视角网络中,在绵阳市中心医院数据集上进行微调,同时利用CBIS-DDSM增加训练的正样本数量以提升网络的泛化能力。实验结果表明,领域适应性学习和数据扩充策略平均提升了17%性能指标,取得了94%和90%的肿块和钙化簇曲线下面积(AUC)值。     

A new feature extraction method based on multi-resolution representations of mammograms

In this paper, I introduce a new method for feature extraction to classify digital mammograms using fast finite shearlet transform. Initially, fast finite shearlet transform was performed over mammogram images, and feature vectors were built using coefficients of the transform. In subsequent calculations, features were ranked according to t-test statistics, and capabilities were distinguished between different classes. To maximize differences between class representatives, a thresholding process was implemented as a final stage of feature extraction, and classifications were calculated over the optimal feature set using 5-fold cross validation and a support vector machine (SVM) classifier. The present results show that the proposed method provides satisfactory classification accuracy.     

基于加权变异免疫算法的微钙化点特征选择

给出了一种乳腺X线照片微钙化点的特征选择方法,该方法运用基于加权变异算子的免疫算法进行特征优选。加权变异算子能够动态调整抗体各部位的变异率,在高亲和力抗体的邻近小范围搜索,在低亲和力抗体的周围跳跃式搜索;为了与支持向量机的分类准则保持一致性,该免疫算法在特征空间中通过核函数计算亲和力。实验使用该方法对微钙化点的20种常用特征进行选择,其结果与经验特征集基本相符但更精简,提高了计算效率,是一种可行的特征选择方法。