Subodh Srivastava

Detection and Classification of Cancer from Microscopic Biopsy Images Using Clinically Significant and Biologically Interpretable Features

A framework for automated detection and classification of cancer from microscopic biopsy images using clinically significant and biologically interpretable features is proposed and examined. The various stages involved in the proposed methodology include enhancement of microscopic images, segmentation of background cells, features extraction, and finally the classification. An appropriate and efficient method is employed in each of the design steps of the proposed framework after making a comparative analysis of commonly used method in each category. For highlighting the details of the tissue and structures, the contrast limited adaptive histogram equalization approach is used. For the segmentation of background cells, k-means segmentation algorithm is used because it performs better in comparison to other commonly used segmentation methods. In feature extraction phase, it is proposed to extract various biologically interpretable and clinically significant shapes as well as morphology based features from the segmented images. These include gray level texture features, color based features, color gray level texture features, Laws Texture Energy based features, Tamuras features, and wavelet features. Finally, the K-nearest neighborhood method is used for classification of images into normal and cancerous categories because it is performing better in comparison to other commonly used methods for this application. The performance of the proposed framework is evaluated using well-known parameters for four fundamental tissues (connective, epithelial, muscular, and nervous) of randomly selected 1000 microscopic biopsy images.

Restoration of Poisson noise corrupted digital images with nonlinear PDE based filters along with the choice of regularization parameter estimation

In this paper, the reconstruction of three nonlinear partial differential equation (PDE) based filters adapted to Poisson noise statistics have been proposed in a variational framework for restoration and enhancement of digital images corrupted with Poisson noise. The proposed and examined PDE based filters include total variation adapted to Poisson noise in L-1 framework; anisotropic diffusion; and complex diffusion based methods adapted to Poisson noise in L-2 framework. The resulting filters contain two terms namely data fidelity and regularization or smoothing function. The data fidelity term is Poisson likelihood term and the regularization functions are PDE based filters. Other choices for the regularization functions have also been presented. The two terms in the proposed filters are coupled with a regularization parameter lambda which makes a proper balance between the two terms during the filtering process. The choice of method for estimation of regularization parameter lambda plays an important role. In this study, the various regularization parameter estimation methods for Poisson noise have also been presented and their suitability has been examined. The resulting optimization problems are further investigated for efficient implementation for large scale problems. For estimating the regularization parameter, three choices are considered for Poisson noise case which are discrepancy principles, generalized cross validations (GCV), and unbiased predictive risk estimate (UPRE). GCV and UPRE functions are further other optimization problems in addition to main image reconstruction problem. For minimizing the GCV and UPRE functions, the methods of Conjugate Gradients (CG) is used. For digital implementations, all schemes have been discretized using finite difference scheme. The comparative analysis of the proposed methods are presented in terms of relative norm error, improvement in SNR, MSE, PSNR, CP and MSSIM for an adaptive value of regularization parameter calculated by every methods in consideration. Finally, from the obtained results it is observed that the anisotropic diffusion based method adapted to Poisson noise gives better results in comparison to other methods in consideration along with choice of GCV for regularization parameter selection.

PDE Based Unsharp Masking, Crispening and High Boost Filtering of Digital Images

A partial differential equation (PDE) based technique is proposed and implemented to perform unsharp masking, crispening and high boost filtering of digital images. The traditional concept of unsharp masking and crispening of edges which uses Laplacian as intermediate step for smoothening the image has been extended and modified using the idea of Perona and Malik [1] which overcomes the disadvantages of Laplacian method. For descretization, finite differences scheme has been used. The scheme has been implemented using MATLAB 7.0 and performance is tested for various gray images of different resolutions and the obtained results justify the applicability of proposed scheme.

Multi-view human activity recognition based on silhouette and uniform rotation invariant local binary patterns

This paper addresses the problem of silhouette-based human activity recognition. Most of the previous work on silhouette based human activity recognition focus on recognition from a single view and ignores the issue of view invariance. In this paper, a system framework has been presented to recognize a view invariant human activity recognition approach that uses both contour-based pose features from silhouettes and uniform rotation local binary patterns for view invariant activity representation. The framework is composed of three consecutive modules: (1) detecting and locating people by background subtraction, (2) combined scale invariant contour-based pose features from silhouettes and uniform rotation invariant local binary patterns (LBP) are extracted, and (3) finally classifying activities of people by Multiclass Support vector machine (SVM) classifier. The rotation invariant nature of uniform LBP provides view invariant recognition of multi-view human activities. We have tested our approach successfully in the indoor and outdoor environment results on four multi-view datasets namely: our own view point dataset, VideoWeb Multi-view dataset [28], i3DPost multi-view dataset [29], and WVU multi-view human action recognition dataset [30]. The experimental results show that the proposed method of multi-view human activity recognition is robust, flexible and efficient.

A combined approach for the enhancement and segmentation of mammograms using modified fuzzy C-means method in wavelet domain.

In this paper, a combined approach for enhancement and segmentation of mammograms is proposed. In preprocessing stage, a contrast limited adaptive histogram equalization (CLAHE) method is applied to obtain the better contrast mammograms. After this, the proposed combined methods are applied. In the first step of the proposed approach, a two dimensional (2D) discrete wavelet transform (DWT) is applied to all the input images. In the second step, a proposed nonlinear complex diffusion based unsharp masking and crispening method is applied on the approximation coefficients of the wavelet transformed images to further highlight the abnormalities such as micro-calcifications, tumours, etc., to reduce the false positives (FPs). Thirdly, a modified fuzzy c-means (FCM) segmentation method is applied on the output of the second step. In the modified FCM method, the mutual information is proposed as a similarity measure in place of conventional Euclidian distance based dissimilarity measure for FCM segmentation. Finally, the inverse 2D-DWT is applied. The efficacy of the proposed unsharp masking and crispening method for image enhancement is evaluated in terms of signal-to-noise ratio (SNR) and that of the proposed segmentation method is evaluated in terms of random index (RI), global consistency error (GCE), and variation of information (VoI). The performance of the proposed segmentation approach is compared with the other commonly used segmentation approaches such as Otsus thresholding, texture based, k-means, and FCM clustering as well as thresholding. From the obtained results, it is observed that the proposed segmentation approach performs better and takes lesser processing time in comparison to the standard FCM and other segmentation methods in consideration.

Design, analysis and classifier evaluation for a CAD tool for breast cancer detection from digital mammograms

In this paper, the design, analysis, and classifier evaluation for a computer aided diagnostics (CAD) tool for early breast cancer detection from mammograms is presented. The design steps of a CAD tool include enhancement, segmentation, feature extraction and selection, and classification of images. A contrast limited histogram equalisation method is used for image enhancement followed by cropping of region of interests. The fuzzy C-means method is used for segmenting abnormalities present in the images. A total of 88 hybrid features are extracted for each image. For feature selection, minimum redundancy and maximum relevancy approach has been used. For decision making, the various classifiers examined for their efficacy, for 322 images available in MIAS database, include SVMs for its various kernel choices, k-NN, and ANN. Finally, it is observed that the SVM classifier for the MLP kernel choice is performing better in comparison to all other classifiers in consideration along with the other design steps ...

Effective mammogram classification based on center symmetric-LBP features in wavelet domain using random forests

Mammogram classification is a crucial and challenging problem, because it helps in early diagnosis of breast cancer and supports radiologists in their decision to analyze similar mammograms out of a database by recognizing the classes of current mammograms. This paper proposes an effective method for classifying mammograms using random forests with wavelet based center-symmetric local binary pattern (WCS-LBP). To classify mammograms, multi-resolution CS-LBP texture characteristics from non-overlapping regions of the mammograms are captured. Further, we examine most relevant features using support vector machine-recursive feature elimination (SVM-RFE). Finally, we feed the selected features to decision trees and construct random forests which are an ensemble of random decision trees. Using wavelet based local CS-LBP features with random forest, we classify the test images into different categories having the maximum posterior probability. The proposed method shows the improved performance as compared with other variant features and state-of-art methods. The obtained performance measures are 97.3% accuracy, 97.3% precision, 97.2% recall, 97.2% F-measure and 94.1% Matthews correlation coefficient (MCC).

An adaptive non-linear PDE-based speckle reduction technique for ultrasound images

In this paper, an adaptive nonlinear complex diffusion based filter is proposed to reduce the speckle noise from ultrasound images. The adaptive and modified version of a fourth order PDE is also proposed and examined for its suitability and efficacy for speckle reduction from ultrasound images. The performance of the proposed method is compared with constraint driven anisotropic diffusion based method, modified fourth order PDE based method and other methods such as Speckle Reducing Anisotropic Diffusion (SRAD) filter, Lee filter, Frost filter and Kuan filter in terms of various performance metrics such as Mean Square Error (MSE), Peak Signal-to-Noise Ratio (PSNR), Correlation Parameter (CP) and Mean Structure Similarity Index Map (MSSIM). The obtained results justify the applicability of the proposed scheme.

An Efficient Image Retrieval Based on Fusion of Fast Features and Query Image Classification

Content Based Image Retrieval (CBIR) is an emerging research area in computer vision, in which, we yield similar images as per the query content. For the implementation of CBIR system, feature extraction plays a vital role, where colour feature is quite remarkable. But, due to unevenly colored or achromatic surfaces, the role of texture is also important. In this paper, an efficient and fast CBIR system is proposed, which is based on a fusion of computationally light weighted colour and texture features; chromaticity moment, colour percentile, and local binary pattern (LBP). Using these features with multiclass classifier, the authors propose a supervised query image classification and retrieval model, which filters all irrelevant class images. Basically, this model categorizes and recovers the class of a query image based on its visual content, and this successful classification of image significantly enhances the performance and searching time of retrieval system. Descriptive experimental analysis on benchmark databases confirms the effectiveness of proposed retrieval framework. KEywoRDS Chromaticity Moment, Classification, Content Based Image Retrieval, Feature Extraction, Similarity Measure

A fourth order PDE based fuzzy c- means approach for segmentation of microscopic biopsy images in presence of Poisson noise for cancer detection

BACKGROUND AND OBJECTIVEnFor cancer detection from microscopic biopsy images, image segmentation step used for segmentation of cells and nuclei play an important role. Accuracy of segmentation approach dominate the final results. Also the microscopic biopsy images have intrinsic Poisson noise and if it is present in the image the segmentation results may not be accurate. The objective is to propose an efficient fuzzy c-means based segmentation approach which can also handle the noise present in the image during the segmentation process itself i.e. noise removal and segmentation is combined in one step.nnnMETHODSnTo address the above issues, in this paper a fourth order partial differential equation (FPDE) based nonlinear filter adapted to Poisson noise with fuzzy c-means segmentation method is proposed. This approach is capable of effectively handling the segmentation problem of blocky artifacts while achieving good tradeoff between Poisson noise removals and edge preservation of the microscopic biopsy images during segmentation process for cancer detection from cells.nnnRESULTSnThe proposed approach is tested on breast cancer microscopic biopsy data set with region of interest (ROI) segmented ground truth images. The microscopic biopsy data set contains 31 benign and 27 malignant images of size 896u2009×u2009768. The region of interest selected ground truth of all 58 images are also available for this data set. Finally, the result obtained from proposed approach is compared with the results of popular segmentation algorithms; fuzzy c-means, color k-means, texture based segmentation, and total variation fuzzy c-means approaches.nnnCONCLUSIONSnThe experimental results shows that proposed approach is providing better results in terms of various performance measures such as Jaccard coefficient, dice index, Tanimoto coefficient, area under curve, accuracy, true positive rate, true negative rate, false positive rate, false negative rate, random index, global consistency error, and variance of information as compared to other segmentation approaches used for cancer detection.