Small bowel image classification using cross-co-occurrence matrices on wavelet domain

This paper presents a novel system to compute the automated classification of wireless capsule endoscope images. Classification is achieved by a classical statistical approach, but novel features are extracted from the wavelet domain and they contain both color and texture information. First, a shift-invariant discrete wavelet transform (SIDWT) is computed to ensure that the multiresolution feature extraction scheme is robust to shifts. The SIDWT expands the signal (in a shift-invariant way) over the basis functions which maximize information. Then cross-co-occurrence matrices of wavelet subbands are calculated and used to extract both texture and color information. Canonical discriminant analysis is utilized to reduce the feature space and then a simple 1D classifier with the leave one out method is used to automatically classify normal and abnormal small bowel images. A classification rate of 94.7% is achieved with a database of 75 images (41 normal and 34 abnormal cases). The high success rate could be attributed to the robust feature set which combines multiresolutional color and texture features, with shift, scale and semi-rotational invariance. This result is very promising and the method could be used in a computer-aided diagnosis system or a content-based image retrieval scheme.     

Tissue counter analysis of histologic sections of melanoma: influence of mask size and shape, feature selection, statistical methods and tissue preparation.

BACKGROUND: Tissue counter analysis is an image analysis tool designed for the detection of structures in complex images at the macroscopic or microscopic scale. As a basic principle, small square or circular measuring masks are randomly placed across the image and image analysis parameters are obtained for each mask. Based on learning sets, statistical classification procedures are generated which facilitate an automated classification of new data sets. OBJECTIVE: To evaluate the influence of the size and shape of the measuring masks as well as the importance of feature selection, statistical procedures and technical preparation of slides on the performance of tissue counter analysis in microscopic images. As main quality measure of the final classification procedure, the percentage of elements that were correctly classified was used. STUDY DESIGN: HE-stained slides of 25 primary cutaneous melanomas were evaluated by tissue counter analysis for the recognition of melanoma elements (section area occupied by tumour cells) in contrast to other tissue elements and background elements. Circular and square measuring masks, various subsets of image analysis features and classification and regression trees compared with linear discriminant analysis as statistical alternatives were used. The percentage of elements that were correctly classified by the various classification procedures was assessed. In order to evaluate the applicability to slides obtained from different laboratories, the best procedure was automatically applied in a test set of another 50 cases of primary melanoma derived from the same laboratory as the learning set and two test sets of 20 cases each derived from two different laboratories, and the measurements of melanoma area in these cases were compared with conventional assessment of vertical tumour thickness. RESULTS: Square measuring masks were slightly superior to circular masks, and larger masks (64 or 128 pixels in diameter) were superior to smaller masks (8 to 32 pixels in diameter). As far as the subsets of image analysis features were concerned, colour features were superior to densitometric and Haralick texture features. Statistical moments of the grey level distribution were of least significance. CART (classification and regression tree) analysis turned out to be superior to linear discriminant analysis. In the best setting, 95% of melanoma tissue elements were correctly recognized. Automated measurement of melanoma area in the independent test sets yielded a correlation of r=0.846 with vertical tumour thickness (p<0.001), similar to the relationship reported for manual measurements. The test sets obtained from different laboratories yielded comparable results. CONCLUSIONS: Large, square measuring masks, colour features and CART analysis provide a useful setting for the automated measurement of melanoma tissue in tissue counter analysis, which can also be used for slides derived from different laboratories.     

基于颜色名和OpponentSIFT特征的鳞翅目昆虫图像识别

【目的】本研究旨在探索使用先进的计算机视觉技术实现对昆虫图像的自动分类方法。【方法】通过预处理对采集的昆虫标本图像去除背景,获得昆虫图像的前景蒙板,并由蒙板确定的轮廓计算出前景图像的最小包围盒,剪切出由最小包围盒确定的前景有效区域,然后对剪切得到的图像进行特征提取。首先提取颜色名特征,把原来的RGB（Red-Green-Blue）图像的像素值映射到11种颜色名空间,其值表示RGB值属于该颜色名的概率,每个颜色名平面划分成3×3像素大小的网格,用每格的概率均值作为网格中心点的描述子,最后用空阈金字塔直方图统计的方式形成颜色名视觉词袋特征;其次提取OpponentSIFT(Opponent Scale Invariant Feature Transform)特征,首先把RGB图像变换到对立色空间,对该空间每通道提取SIFT特征,最后用空域池化和直方图统计方法形成OpponentSIFT视觉词袋。将两种词袋特征串接后得到该昆虫图像的特征向量。使用昆虫图像样本训练集提取到的特征向量训练SVM（Support Vector Machine）分类器,使用这些训练得到的分类器即可实现对鳞翅目昆虫的分类识别。【结果】该方法在包含10种576个样本的昆虫图像数据库中进行了测试,取得了100%的识别正确率。【结论】试验结果证明基于颜色名和OpponentSIFT特征可以有效实现对鳞翅目昆虫图像的识别。     

Segmentation of stained blood cell images measured at high scanning density with high magnification and high numerical aperture optics

In hematological morphology, it is necessary to resolve and analyze the smallest possible cellular details appearing in the light microscope. A prerequisite for computer-aided analysis of subtle morphological features is measuring the cells at a high scanning density with high magnification and high numerical aperture optics. Contrary to visual observations, the information content in a measured picture can be increased by setting the condensor's numerical aperture (NA) greater than the objective's NA. The complexity and heterogeneity of such cell images necessitate a new segmentation method that conserves the morphological information required in the subsequent image analysis, feature extraction, and cell classification. In our segmentation strategy, characteristic color difference thresholds for each nucleus and cytoplasm are combined with geometric operations, probability functions, and a cell model. All thresholds are repeatedly recalculated during the successive improvements of the image masks. None of the thresholds are fixed. This strategy segments blood cell images containing touching cells and large variations in staining, texture, size, and shape. Biological inconsistencies in the calculated cell masks are eliminated by comparing each mask with the cell model criteria integrated into the entire segmentation process. All 20,000 leukocyte images from 120 smears in our leukemia project were segmented with this method.     

High-resolution and contextual analysis for the diagnosis of fine needle aspirates of breast.

A study was undertaken to confirm earlier work on a smaller number of patients that had suggested that medium-resolution contextual analysis complements high-resolution individual cell analysis for cytomorphometric classification of fine needle aspirate smears of breast. The objectives of this study were to improve and verify the method. Sixty-one biopsy-confirmed hematoxylin and eosin-stained aspirate smears of breast were restained using the Feulgen technique. Individual nuclei were digitized at a resolution of 0.25 micron. Features describing size, shape, density and texture were extracted from the images. Individual cell analysis correctly classified 84% of cases, contextual analysis correctly classified 70% of cases, and the combined use of both techniques resulted in 87% classification accuracy. However, if fibroadenoma cases are excluded, the combined correct classification rate is 93%. Geometric and densitometric features contributed most to correct classification in individual cell analysis, while the most important contextual feature was the number of clusters per scene. We conclude that the addition of quantitative measures of smear patterns, termed "contextual analysis," improves automated classification schemes.     

Direct identification of pure Penicillium species using image analysis

This paper presents a method for direct identification of fungal species solely by means of digital image analysis of colonies as seen after growth on a standard medium. The method described is completely automated and hence objective once digital images of the reference fungi have been established. Using a digital image it is possible to extract precise information from the surface of the fungal colony. This includes color distribution, colony dimensions and texture measurements. For fungal identification, this is normally done by visual observation that often results in a very subjective data recording. Isolates of nine different species of the genus Penicillium have been selected for the purpose. After incubation for 7 days, the fungal colonies are digitized using a very accurate digital camera. Prior to the image analysis each image is corrected for self-illumination, thereby gaining a set of directly corresponding images with respect to illumination. A Windows application has been developed to locate the position and size of up to three colonies in the digitized image. Using the estimated positions and sizes of the colonies, a number of relevant features can be extracted for further analysis. The method used to determine the position of the colonies will be covered as well as the feature selection. The texture measurements of colonies of the nine species were analyzed and a clustering of the data into the correct species was confirmed. This indicates that it is indeed possible to identify a given colony merely by macromorphological features. A classifier (in the normal distribution) based on measurements of 151 colonies incubated on yeast extract sucrose agar (YES) was used to discriminate between the species. This resulted in a correct classification rate of 100% when used on the training set and 96% using cross-validation. The same methods applied to 194 colonies incubated on Czapek yeast extract agar (CYA) resulted in a correct classification rate of 98% on the training set and 71% using cross-validation.     

基于复合叶片特征的计算机植物识别方法

该文探讨如何根据植物的叶片特征,利用图像处理和机器学习的方法对植物进行分类。鉴于现有的叶片分类系统多采用单一的特征,如几何和纹理等,仅能在小规模数据库上得到较好的结果。然而,随着样本种类的增多,单一特征在不同种类叶片之间的相似性非常明显,致使分类正确率降低。该研究使用多种复合特征,并提出了原创的预处理方法以及宽度、叶缘频率特征,较传统的几何特征更为详尽。研究结果显示,复合特征可以有效避免算法过拟合问题,使之适用于更大的数据库。通过提取21类植物的叶片宽度、颜色、叶缘和纹理共292维特征,对1 915张数字图像进行了分类,正确率达到93%,并分析了各类特征对分类结果的影响。研究结果表明,在不影响分类正确率前提下,可将特征减少到约100维。     

Scene-segmentation algorithm development using error measures

Development of scene-segmentation algorithms has generally been an ad hoc process. This paper presents a systematic technique for developing these algorithms using error-measure minimization. If scene segmentation is regarded as a problem of pixel classification whereby each pixel of a scene is assigned to a particular object class, development of a scene-segmentation algorithm becomes primarily a process of feature selection. In this study, four methods of feature selection were used to develop segmentation techniques for cervical cytology images: (1) random selection, (2) manual selection (best features in the subjective judgment of the investigator), (3) eigenvector selection (ranking features according to the largest contribution to each eigenvector of the feature covariance matrix) and (4) selection using the scene-segmentation error measure A2. Four features were selected by each method from a universe of 35 features consisting of gray level, color, texture and special pixel neighborhood features in 40 cervical cytology images . Evaluation of the results was done with a composite of the scene-segmentation error measure A2, which depends on the percentage of scenes with measurable error, the agreement of pixel class proportions, the agreement of number of objects for each pixel class and the distance of each misclassified pixel to the nearest pixel of the misclassified class. Results indicate that random and eigenvector feature selection were the poorest methods, manual feature selection somewhat better and error-measure feature selection best. The error-measure feature selection method provides a useful, systematic method of developing and evaluating scene-segmentation algorithms.     

The rejection of noncellular artifacts in Papanicolaou-stained slide specimens by an automated high-resolution system. Identification of important cytometric features

The important cytodiagnostic features that permit discrimination of typical cell types by high-resolution image analysis and pattern recognition techniques have been previously studied in detail. An automated system for the diagnosis of Papanicolaou-stained specimens must also deal, however, with the "real world" of extraneous noncellular artifacts and debris found on every slide. Features that are ideal for the separation of typical normal and abnormal cells may not be adequate by themselves to reject these objects. A new set of discriminatory features must be found. In order to identify those features, a large set of images acquired using the TICAS high-resolution television rapid-scanning system was analyzed and studied. These images, from a variety of slide types, included normal cells, abnormal cells and noncellular artifacts identified by low-resolution preprocessing logic as suspicious enough to warrant high-resolution study. The results indicate that the more important features for such discrimination are not those traditionally important in distinguishing abnormal from normal cells but include color relations, shape measures, boundary properties and texture features.     

Classification of lymphoproliferative disorders by spectral imaging of the nucleus

Spectral nuclear morphometry was used for the classification of lymphocytes in lymphoproliferative disorders. May-Grunwald-Giemsa-stained blood specimens were taken from thirty patients with infectious mononucleosis, non-Hodgkin lymphoma or chronic lymphocytic leukemia, and from ten healthy individuals. Blood specimens were analyzed by spectral imaging. Seventeen distinct spectra were collected into a spectral library and a distinct pseudo color was assigned to each one of them. The library was used to scan all the cells in the database and to create a spectrally classified image of each cell. The spectral map, per cell, reveals distinct spectral-response regions in each cellular compartment, via the distinct region colors. Computational analysis of the spectral maps allows for the objective quantification of a set of parameters, or features, representing the cell. The features used in this work include the area and perimeter of the nucleus, circularity, edginess and the spectral pattern. The analysis pursued showed that each class of cells is associated with a set of unique parameters. We conclude that spectral analysis combined with feature analysis provides significant information in the analysis of lymphoproliferative disorders and may serve as an additional tool for the histopathological evaluation of disease.     

Performance of Feature Selection Methods

High-throughput biological technologies offer the promise of finding feature sets to serve as biomarkers for medical applications; however, the sheer number of potential features (genes, proteins, etc.) means that there needs to be massive feature selection, far greater than that envisioned in the classical literature. This paper considers performance analysis for feature-selection algorithms from two fundamental perspectives: How does the classification accuracy achieved with a selected feature set compare to the accuracy when the best feature set is used and what is the optimal number of features that should be used? The criteria manifest themselves in several issues that need to be considered when examining the efficacy of a feature-selection algorithm: (1) the correlation between the classifier errors for the selected feature set and the theoretically best feature set; (2) the regressions of the aforementioned errors upon one another; (3) the peaking phenomenon, that is, the effect of sample size on feature selection; and (4) the analysis of feature selection in the framework of high-dimensional models corresponding to high-throughput data.     

水果形状的傅里叶描述子研究

水果的形状是水果分级的重要指标之一。本文研究了不规则物体形状的数学描述方法,认为在水果的分级过程中采用曲线拟合的方法来描述水果的形状是不合适的;提出了仅需利用物体的边界信息求物体的形心坐标和描述果形的新方法;发现用Fourier描述子的前4个谐波分量的变化特性就能较好地代表水果的形状,用前15个谐波分量来描述形状则可以达到相当高的精度。而且傅立叶描述子可以进行平移、旋转和缩放,并具有很强的水果外形重建功能,是一描述水果形状的非常有效的方法。