Classification Of Microscopic Images Research Articles

A characteristic absorption peak interval method based on subspace partition for FTIR microscopic imaging classification

A spatial distribution of an image that retains all of the multicomponent sample information in the spectral channels can be obtained using multivariate analysis methods, such as principal component analysis (PCA). Most multivariate methods build classifiers based on the spectral features extracted from high-dimensional space. However, such mathematical models have been devoted to exploring the spectral features contained in full spectrum bands, and lack the chemical specificity of the mid-infrared spectrum. In this report, we present the results of a novel characteristic absorption peak interval (CAPI) method to extract spectral band characteristics. This CAPI method extracts absorption peak bands from the spectral dimension by implementing four developed strategies of subspace partition (SP), thereby capturing the subspace characteristic information in multiple adjacent functional group areas. Finally, stacked absorption peak bands are utilized to obtain more efficient distribution information from the multiple subspace feature sets, and then an extreme learning machine is used to perform classification. Experimental results show that the proposed all-sub-band randomization strategy, subspace randomization strategy, probabilistic principal component analysis of the subspace partition, and optimum index factor of the subspace partition are all effective for mid-infrared spectroscopy microscopic image classification. Compared to PCA, 2D principal component analysis, probabilistic principal component analysis, kernel principal component analysis, and PCA linear discriminant analysis, experimental results for three Fourier transform infrared spectroscopy (FTIR) microscopy imaging datasets show that the proposed CAPI method outperforms the other multivariate methods, achieves a higher overall and average classification accuracy, and retains the physical meaning of the spectrum.

A Proposed Novel Method for Detection and Classification of Leukemia using Blood Microscopic Images

Leukemia is a form of cancer in the blood cells. Most forms of Leukemia occur in the White Blood Cells. The aim of the study is to recognize leukemia White Blood Cells from blood microscopic images through segmentation. The proposed work, an Enhanced Hybrid Fuzzy C-Means with Cluster Center Estimation is to separate the nucleus from the White Blood Cell image which is received as an output of the Hybrid Fuzzy C-Means with Cluster Center Estimation algorithm. The structural image modelling technique called morphological operation is utilized to extract the nucleus followed by feature extraction to extract the geometric features of the nuclei. The Support Vector Machine which is an empirically good performance pattern recognition tool to perform the classification and prediction of cancer affected blood cell images. An analysis is carried by considering the extracted feature of White Blood Cell and nucleus. Finally the performance of this method is evaluated on the basis of image quality measures such as Peak Signal to Noise Ratio and Mean Square Error. Thus, this method results in a high Peak Signal to Noise Ratio and low Mean Square Error, which indicates the better reconstruction of the images in comparison with other existing methods. Keywords: Leukemia, Segmentation, Morphology, Support Vector Machine, Peak Signal to Noise Ratio, Mean Square Error

A transfer learning method for automatic identification of sandstone microscopic images

Classification of sandstone microscopic images is an essential task in geology, and the classical method is either subjective or time-consuming. Computer aided automatic classification has been proved useful, but it seldom considers the situation where sandstone images are collected from separated regions. In this paper, we provide a method called Festra, which uses transfer learning to handle the problem of interregional sandstone microscopic image classification. The method contains two parts: one is feature selection, which aims to screen out features having great difference between the regions, the other is instance transfer using an enhanced TrAdaBoost, whose object is to mitigate the difference among thin section images collected from the regions. Experiments are conducted based on the sandstone images taken from four regions in Tibet to study the performance of Festra. The experimental results have proved both effectiveness and validity of Festra, which provides competitive prediction performance on all the four regions, with few target instances labeled suitable for the field use.

Binary wavelet transform-based completed local binary pattern texture descriptors for classification of microscopic images of hardwood species

This paper presents an approach for generating a binary wavelet transform-based completed local binary pattern (BWTCLBP) texture descriptor to improve the classification accuracy of microscopic images of hardwood species. Firstly, gray-level slicing method is used to obtain eight (b0–b7) bit planes from grayscale image. Then, the two-dimensional binary wavelet transform (2D-BWT) decomposes each of the most significant bit-plane (b7) images up to the fifth scale of decomposition. The texture descriptors are then acquired from each of the subimages up to the five scales of decomposition. Further, two variants of support vector machine (SVM), linear SVM and radial basis function kernel SVM, were employed as classifiers. The classification accuracy of the proposed and existing texture descriptors was compared. The BWT-based uniform completed local binary pattern (BWTCLBPu2) texture descriptor achieved the best classification accuracy of 95.07 ± 0.72% at the third scale of decomposition. The classification accuracy is produced by linear SVM classifier for full feature (1416) vector data. In order to overcome the effect of curse of dimensionality, the minimal-redundancy–maximal-relevance feature selection method is employed to select the best subset of feature vector data. This approach has resulted in improved classification accuracy of 96.60 ± 0.80% (450) by linear SVM classifier.

Nonparametric de-noising filter optimization using structure-based microscopic image classification.

The Local Polynomial Approximation (LPA) is a nonparametric filter that performs pixel-wise polynomial fit on a certain neighborhood. This filter can be supported by the Intersection of Confidence Interval rule (ICI) as an adaptation algorithm to identify the most suited neighborhood at which the polynomial assumptions provide superior fit for the observations. However, the LPA-ICI is considered to be a near-optimal de-noising filter. Moreover, the ICI rule has several parameters that affect its performance. The current study applied an optimization algorithm, namely the Particle swarm optimization (PSO) to determine the optimal ICI parameter values for microscopic images de-noising. As the ICI parameters are image's structure based, bag-of-features classifier is used to classify the images based on their structure into different classes. Afterward, a generated optimal ICI parameters' table was created using the LPA-ICI-PSO for further direct use without optimization. This table included the optimal ICI parameters based on the image structure. Based on the image category, the generated table can be used to attain the suitable ICI optimal parameters without using PSO. This guarantees less computational time along with the optimal de-noising compared to the LPA-ICI as established by the performance metrics. The experimental results established the superiority of the proposed LPA-ICI-PSO over the classical LPA-ICI filter.

Microscopic medical image classification framework via deep learning and shearlet transform.

Cancer is the second leading cause of death in US after cardiovascular disease. Image-based computer-aided diagnosis can assist physicians to efficiently diagnose cancers in early stages. Existing computer-aided algorithms use hand-crafted features such as wavelet coefficients, co-occurrence matrix features, and recently, histogram of shearlet coefficients for classification of cancerous tissues and cells in images. These hand-crafted features often lack generalizability since every cancerous tissue and cell has a specific texture, structure, and shape. An alternative approach is to use convolutional neural networks (CNNs) to learn the most appropriate feature abstractions directly from the data and handle the limitations of hand-crafted features. A framework for breast cancer detection and prostate Gleason grading using CNN trained on images along with the magnitude and phase of shearlet coefficients is presented. Particularly, we apply shearlet transform on images and extract the magnitude and phase of shearlet coefficients. Then we feed shearlet features along with the original images to our CNN consisting of multiple layers of convolution, max pooling, and fully connected layers. Our experiments show that using the magnitude and phase of shearlet coefficients as extra information to the network can improve the accuracy of detection and generalize better compared to the state-of-the-art methods that rely on hand-crafted features. This study expands the application of deep neural networks into the field of medical image analysis, which is a difficult domain considering the limited medical data available for such analysis.

Morphometric information to reduce the semantic gap in the characterization of microscopic images of thyroid nodules

BackgroundThe analyses of several systems for medical-imaging processing typically support the extraction of image attributes, but do not comprise some information that characterizes images. For example, morphometry can be applied to find new information about the visual content of an image. The extension of information may result in knowledge. Subsequently, results of mappings can be applied to recognize exam patterns, thus improving the accuracy of image retrieval and allowing a better interpretation of exam results. Although successfully applied in breast lesion images, the morphometric approach is still poorly explored in thyroid lesions due to the high subjectivity thyroid examinations. ObjectiveThis paper presents a theoretical–practical study, considering Computer Aided Diagnosis (CAD) and Morphometry, to reduce the semantic discontinuity between medical image features and human interpretation of image content. MethodThe proposed method aggregates the content of microscopic images characterized by morphometric information and other image attributes extracted by traditional object extraction algorithms. This method carries out segmentation, feature extraction, image labeling and classification. Morphometric analysis was included as an object extraction method in order to verify the improvement of its accuracy for automatic classification of microscopic images. ResultsTo validate this proposal and verify the utility of morphometric information to characterize thyroid images, a CAD system was created to classify real thyroid image-exams into Papillary Cancer, Goiter and Non-Cancer. Results showed that morphometric information can improve the accuracy and precision of image retrieval and the interpretation of results in computer-aided diagnosis. For example, in the scenario where all the extractors are combined with the morphometric information, the CAD system had its best performance (70% of precision in Papillary cases). ConclusionResults signalized a positive use of morphometric information from images to reduce semantic discontinuity between human interpretation and image characterization.

Gaussian image pyramid based texture features for classification of microscopic images of hardwood species

This paper presents a texture feature based approach for hardwood species classification. The three existing feature extraction techniques such as local binary pattern (LBP), local configuration pattern (LCP) and local phase quantization (LPQ) are integrated here with Gaussian image pyramid (GIP) which results in improvement of classification accuracy. The texture features are extracted at seven different decomposition levels generated by the GIP. These texture features are fed as input to linear support vector machine (SVM) classifier that uses 10-fold cross validation approach of classification. The results of combination of GIP decomposition with individual texture feature extraction techniques and linear SVM classifier have been compared. The comparison yields that Gaussian image pyramid based local phase quantization (GPLPQ) texture feature extraction technique using third (3rd) level of image decomposition results in the best classification accuracy of 98.60% for hardwood species. The proposed integration of GIP and texture feature extraction techniques also proves to be an effective tool of classification for texture surface database. For texture surface database, Gaussian image pyramid based rotation invariant uniform local configuration pattern (GPLCPriu2) has achieved 98.00% classification accuracy.

Multiresolution local binary pattern variants based texture feature extraction techniques for efficient classification of microscopic images of hardwood species

In this paper, multiresolution local binary pattern (MRLBP) variants based texture feature extraction techniques have been proposed to categorize hardwood species into its various classes. Initially, discrete wavelet transform (DWT) has been used to decompose each image up to 7 levels using Daubechies wavelet (db2) as decomposition filter. Subsequently, six texture feature extraction techniques (local binary pattern and its variants) are employed to obtain substantial features of these images at different levels. Three classifiers, namely, linear discriminant analysis (LDA), linear and radial basis function (RBF) kernel support vector machine (SVM), have been used to classify the images of hardwood species. Thereafter, classification results obtained from conventional and MRLBP variants based texture feature extraction techniques with different classifiers have been compared. For 10-fold cross validation approach, texture features acquired using discrete wavelet transform based uniform completed local binary pattern (DWTCLBPu2) feature extraction technique has produced best classification accuracy of 97.40±1.06% with linear SVM classifier. This classification accuracy has been achieved at the 3rd level of image decomposition using full feature (1416) dataset. Further, reduction in dimension of texture features (325 features) by principal component analysis (PCA) has been done and the best classification accuracy of 97.87±0.82% for DWTCLBPu2 at the 3rd level of image decomposition has been obtained using LDA classifier. The DWTCLBPu2 texture features have also established superiority among the MRLBP techniques with reduced dimension features for randomly divided database into fix training and testing ratios.

Implementation of Wavelets and Artificial Neural Networks in Colonic Histopathological Classification

Colon cancer is one type of cancer that has a high death rate, but early diagnosis can improve the chances of patient recovery. Computer-assisted diagnosis can aid in determining whether images are of healthy or cancerous tissues. This study aims to contribute to the automatic classification of microscopic colonic images by implementing a 2-D wavelet transform for feature extraction and neural networks for classification. The colonic histopathological images are assigned to either the normal, cancerous, or adenomatous polyp classes. The proposed algorithm is able to determine which of the three classes the images belong to at a 91.11% rate of accuracy.

Multiple Ranks Weighting Score for Microscopic Image Retrieval System

Content based medical images have become a major necessity with the growing retrieving Advancements. CBIR access to medical images for supporting clinical decision making has been proposed that would be ease to manage large number of image in the database system. (4) In real time case only few systems has been developed and used in clinical environment. Content-Based Image Retrieval refers to image retrieval system that is based on visual properties of image objects other than textual annotation. Query image features compare with the database image features which is not exactly matching so image feature can be compare with the two tier approach in the database image in order to improve the accuracy of the retrieval system. Every day, large volume of different types of medical images such as MRI, CT images ultrasound, x- ray, radiology, etc are produced in different medical centre's .microscopic image classification and discrimination (sub-type) (12) is the most difficult problem in medical image retrieval system. In this paper, the survey provides the suitable algorithm for retrieval and classification of medical image to improve the overall accuracy of the MIMS.

A New Method Based on Adaptive Discrete Wavelet Entropy Energy and Neural Network Classifier (ADWEENN) for Recognition of Urine Cells from Microscopic Images Independent of Rotation and Scaling

So far, analysis and classification of urine cells number has become an important topic for medical diagnosis of some diseases. Therefore, in this study, we suggest a new technique based on Adaptive Discrete Wavelet Entropy Energy and Neural Network Classifier (ADWEENN) for Recognition of Urine Cells from Microscopic Images Independent of Rotation and Scaling. Some digital image processing methods such as noise reduction, contrast enhancement, segmentation, and morphological process are used for feature extraction stage of this ADWEENN in this study. Nowadays, the image processing and pattern recognition topics have come into prominence. The image processing concludes operation and design of systems that recognize patterns in data sets. In the past years, very difficulty in classification of microscopic images was the deficiency of enough methods to characterize. Lately, it is seen that, multi-resolution image analysis methods such as Gabor filters, discrete wavelet decompositions are superior to other classic methods for analysis of these microscopic images. In this study, the structure of the ADWEENN method composes of four stages. These are preprocessing stage, feature extraction stage, classification stage and testing stage. The Discrete Wavelet Transform (DWT) and adaptive wavelet entropy and energy is used for adaptive feature extraction in feature extraction stage to strengthen the premium features of the Artificial Neural Network (ANN) classifier in this study. Efficiency of the developed ADWEENN method was tested showing that an avarage of 97.58% recognition succes was obtained.

The application of pattern recognition in the automatic classification of microscopic rock images

The classification of rocks is an inherent part of modern geology. The manual identification of rock samples is a time-consuming process, and—due to the subjective nature of human judgement—burdened with risk. In the course of the study discussed in the present paper, the authors investigated the possibility of automating this process.During the study, nine different rock samples were used. Their digital images were obtained from thin sections, with a polarizing microscope. These photographs were subsequently classified in an automatic manner, by means of four pattern recognition methods: the nearest neighbor algorithm, the K-nearest neighbor, the nearest mode algorithm, and the method of optimal spherical neighborhoods. The effectiveness of these methods was tested in four different color spaces: RGB, CIELab, YIQ, and HSV.The results of the study show that the automatic recognition of the discussed rock types is possible. The study also revealed that, if the CIELab color space and the nearest neighbor classification method are used, the rock samples in question are classified correctly, with the recognition levels of 99.8%.

A new method for automated detection and identification of neurons in microscopic images of brain slices

A new combined mathematical method is proposed, implemented, and experimentally tested for extracting information necessary for modeling and, in future, predicting Parkinson's disease (PD) from microscopic images of brain slices of experimental animals. The method allows one to detect and identify as neurons a set of small informative extended objects with well distinguished (by brightness) oval inclusions. The result is a binary image of the contours of detected objects and their inclusions and a list of characteristics calculated for each detected object. The method is based on the joint application of image processing methods, methods of mathematical morphology, methods of segmentation, and the methods of classification of microscopic images.

Breast cancer diagnosis from biopsy images with highly reliable random subspace classifier ensembles

Accurate and reliable classification of microscopic biopsy images is an important issue in computer assisted breast cancer diagnosis. In this paper, a new cascade Random Subspace ensembles scheme with reject options is proposed for microscopic biopsy image classification. The classification system is built as a serial fusion of two different Random Subspace classifier ensembles with rejection options to enhance the classification reliability. The first ensemble consists of a set of Support Vector Machine classifiers that converts the original $$K$$ -class classification problem into a number of $$K$$ 2-class problems. The second ensemble consists of a Multi-Layer Perceptron ensemble, that focuses on the rejected samples from the first ensemble. For both of the ensembles, the reject option is implemented by relating the consensus degree from majority voting to a confidence measure, and abstaining to classify ambiguous samples if the consensus degree is lower than some threshold. We also investigated the effectiveness of a feature description approach by combining Local Binary Pattern (LBP) texture analysis, statistics derived using the Gray Level Co-occurrence Matrix (GLCM) and the Curvelet Transform. While the LBP analysis efficiently describes local texture properties and the GLCM reflects global texture statistics, the Curvelet Transform is particularly appropriate for the representation of piece-wise smooth images with rich edge information. The combined feature description thus provides a comprehensive biopsy image characterization by taking advantages of their complementary strengths. Using a benchmark microscopic biopsy image dataset, obtained from the Israel Institute of Technology, a high classification accuracy of $$99.25 \%$$ was obtained (with a rejection rate of $$1.94 \%$$ ) using the proposed system.

Fuzzy Rule based Classification of Microscopic Images of Squamous Cell Carcinoma of Esophagus

paper, we present a novel method based on active contours for segmentation and fuzzy rule based classification of microscopic images of esophagus tissues obtained from the abnormal regions of human esophagus detected through endoscopy. This method is used for classification of Squamous Cell Carcinoma (SCC) of esophagus, namely, well differentiated (WD), moderately differentiated (MD), and poorly differentiated (PD) SCC. The multi-grid active contour method is used for cell nuclei segmentation, three geometric features are used feature extraction and a fuzzy rule based classifier is built for classification of SCC. The experimental results demonstrate the efficacy of the proposed method. Keywordsmicroscopic image, squamous cell carcinoma, geometric features, active contour segmentation

Rational Variety Mapping for Contrast-Enhanced Nonlinear Unsupervised Segmentation of Multispectral Images of Unstained Specimen

A methodology is proposed for nonlinear contrast-enhanced unsupervised segmentation of multispectral (color) microscopy images of principally unstained specimens. The methodology exploits spectral diversity and spatial sparseness to find anatomical differences between materials (cells, nuclei, and background) present in the image. It consists of rth-order rational variety mapping (RVM) followed by matrix/tensor factorization. Sparseness constraint implies duality between nonlinear unsupervised segmentation and multiclass pattern assignment problems. Classes not linearly separable in the original input space become separable with high probability in the higher-dimensional mapped space. Hence, RVM mapping has two advantages: it takes implicitly into account nonlinearities present in the image (ie, they are not required to be known) and it increases spectral diversity (ie, contrast) between materials, due to increased dimensionality of the mapped space. This is expected to improve performance of systems for automated classification and analysis of microscopic histopathological images. The methodology was validated using RVM of the second and third orders of the experimental multispectral microscopy images of unstained sciatic nerve fibers (nervus ischiadicus) and of unstained white pulp in the spleen tissue, compared with a manually defined ground truth labeled by two trained pathophysiologists. The methodology can also be useful for additional contrast enhancement of images of stained specimens.

Telling one from another

IT is still an open question that how images of individual cells can be identified automatically in an unsupervised manner. Furthermore, it is a great challenge to quantify cell morphology and intracellular structures either in three or even four dimensions given the great heterogeneity between cells and the heterogeneity in quality between images and samples. These important aspects of a quantitative and unbiased cell analysis for cytomics or cell systems analysis have been frequent issues in the recent past (1,2). Particularly, neurite morphology is very complex with a number of features relevant in information transduction and processing, such as the number of dendrites, their length, branching and branching points, number and type of synapses, or axon length (3). A method of quantitative analysis commonly used in neuronal studies to characterize the morphological characteristics of an imaged neuron—first used in 1953 to describe the differences in the visual and motor cortices of cats—involves initial quantification of a neuron. The analysis is performed by counting the number of dendrite intersections for concentric circles usually centered at the centroid of the cell body, of gradually increasing radius. This approach is now further elucidated and refined by Langhammer and coworkers (this issue, page 1160) using the Sholl analysis of arbor subregions of the dendrites, leading to new biological findings. Separating automatically or interactively close-by objects such as cells in confluent cell or tissue cultures can be a demanding task, even for an experienced observer, not to mention a digital one (4). Like a drop of water falling on a topographic relief flows along a path, such is the grey-shade of a microscopic image of cells, to finally reach a local minimum; the watershed of a relief corresponds to the limits of the adjacent minimum between imaged objects. These algorithms are termed Watershed algorithms and are in image analysis, applied to separate close-by objects. This can be applied for single images or three-dimensional image stacks (5). Now, Kachouie and coworkers (this issue, page 1148) propose a constrained watershed tool by which they can segment up to 99% of cells in confluent cultures, a method that also works for mosaic images. Du and coworkers (this issue, page 1137) use an alternative approach to split objects with ill-defined boundaries. This method is interactive and applies geodesic commute distance: a combination of the geodesic and the commute distance approach. This method outperforms presently applied segmentation methods identifying cell images and splitting them even if signals are weak and cells are overlapping. In his Communication to the Editor, Metze (this issue, page 1101) discusses some aspects of the recently published work by Wang et al. (6). This work focused on classification of follicular lesions of the thyroid. The authors describe a computerized method to detect and classify follicular adenoma of the thyroid, follicular carcinoma of the thyroid, and normal thyroid based on the nuclear chromatin distribution from digital images of tissue obtained by routine histological methods. In conclusion, the authors state that nuclear structure alone contains enough information to automatically classify the normal thyroid, follicular carcinoma, and follicular adenoma, as long as groups of nuclei (instead of individual ones) are used. Metze adds an interesting and important aspect to this work, namely that the lesion size is a critical and a potentially determinative aspect of correct lesion classification. Although the aforementioned articles have the automated recognition, segmentation, and classification of microscopic images on the micro (lm) scale as the major focus, two other studies address image manipulation or pattern recognition on the macro scale. Lahrmann and coworkers (this issue, page 1169) addressed the problem of correctly assigning tissue sample images from tissue micro arrays (TMAs) based on whole slide imaging data. TMAs have only been in use since not much more than 10 years. They are very valuable tools to test the binding of drugs to different rare biopsies in high-

Image classification of microscopic colonic images using textural properties and KSOM

Colorectal cancer is considered the third most common neoplasm in the world. Traditionally, pathologists use a microscope to examine histopathological images of biopsy samples taken from patients and make judgments based on their professional expertise. Since this procedure is performed by a human expert, it is therefore subject to inconsistencies due to factors that might affect human performance. To overcome this problem, this paper proposes the use of Kohonen self-organising map and Haralick texture in the analysis of microscopic colonic images. The results presented here are preliminary and show great promise.

Radial Basis Function Neural Networks Classification for the Recognition of Idiopathic Pulmonary Fibrosis in Microscopic Images

This study investigates the potential of applying the radial basis function (RBF) neural network architecture for the classification of biological microscopic images displaying lung tissue sections with idiopathic pulmonary fibrosis. For the development of the RBF classifiers, the fuzzy means clustering algorithm is utilized. This method is based on a fuzzy partition of the input space and requires only a short amount of time to select both the structure and the parameters of the RBF classifier. The new technique was applied in lung sections acquired using a microscope and captured by a digital camera, at a magnification of 4 x. Age- and sex-matched, 6- to 8-week-old mice (five for each time point and five as control) were used for the induction of pulmonary fibrosis (cf. bleomycin). Bleomycin administration initially induces lung inflammation that is followed by a progressive destruction of the normal lung architecture. The captured images correspond to 7, 15, and 23 days after bleomycin or saline injection and bronchoalveolar lavage (BAL) has been performed to the mice sample. The images were analyzed and color features were extracted. A support vector machines (SVMs)-based classifier was also employed for the same problem. The resulting scores derived by visual assessment of the images by expert pathologists were compared with the RBF and SVM classification outcome. Overall, the RBF neural network had a slightly better performance than that of the SVM classifier, but both performed very well, matching to a great percentage the scoring of the experts. There are some erroneous predictions of the algorithm for the regions characterized as "ill" regions (i.e., some bronchia were wrongly classified as fibrotic areas); however, in general, the algorithm worked pretty fine in distinguishing pathologic from normal in most cases and for heterogeneous fibrotic foci, achieving high values in terms of specificity and sensitivity.