Madhumala Ghosh

Automated leukocyte recognition using fuzzy divergence.

This paper aims at introducing an automated approach to leukocyte recognition using fuzzy divergence and modified thresholding techniques. The recognition is done through the segmentation of nuclei where Gamma, Gaussian and Cauchy type of fuzzy membership functions are studied for the image pixels. It is in fact found that Cauchy leads better segmentation as compared to others. In addition, image thresholding is modified for better recognition. Results are studied and discussed.

Machine learning approach for automated screening of malaria parasite using light microscopic images

The aim of this paper is to address the development of computer assisted malaria parasite characterization and classification using machine learning approach based on light microscopic images of peripheral blood smears. In doing this, microscopic image acquisition from stained slides, illumination correction and noise reduction, erythrocyte segmentation, feature extraction, feature selection and finally classification of different stages of malaria (Plasmodium vivax and Plasmodium falciparum) have been investigated. The erythrocytes are segmented using marker controlled watershed transformation and subsequently total ninety six features describing shape-size and texture of erythrocytes are extracted in respect to the parasitemia infected versus non-infected cells. Ninety four features are found to be statistically significant in discriminating six classes. Here a feature selection-cum-classification scheme has been devised by combining F-statistic, statistical learning techniques i.e., Bayesian learning and support vector machine (SVM) in order to provide the higher classification accuracy using best set of discriminating features. Results show that Bayesian approach provides the highest accuracy i.e., 84% for malaria classification by selecting 19 most significant features while SVM provides highest accuracy i.e., 83.5% with 9 most significant features. Finally, the performance of these two classifiers under feature selection framework has been compared toward malaria parasite classification.

Statistical pattern analysis of white blood cell nuclei morphometry

Quantitative microscopy has strengthened conventional diagnostic scheme through better understanding of microscopic features from clinical perspective. Towards this, pathological image analysis has gained immense significance among medical fraternity through visualization and quantitative evaluation of clinical features. Till today pathological inspection of human blood is solely dependent on subjective assessment which usually leads to significant inter-observer variation in grading and subsequently resulting in late diagnosis of certain disease. This paper introduced a systematic approach to morphologically characterize five types of white blood cells (WBC) through statistical pattern analytics. Marker controlled watershed segmentation embedded with morphological operator is employed to segment WBC and its nuclei from light microscopic image of blood samples. Henceforth, one cellular and eight nuclei-based geometric features are computed mathematically and analyzed statistically with t-test and kernel density functions to show their discriminating potentiality among the groups. Amongst all these features, only four statistical significant features are fed to Naïve Bayes classifier for pattern identification with 83.2% overall accuracy. Detailed results are also given here.

Invariant moment based feature analysis for abnormal erythrocyte recognition

Erythrocyte shape recognition is very important in the detection of thalassemia and anemia using microscopic images. This study aims to develop a computer aided shape recognizer for the recognition of abnormal shapes viz., tear drop, echinocyte, eliptocyte. Here such recognition is done using Hus moments and other geometric features followed by gray level thresholding and marker controlled watershed segmentation. These features are statistically evaluated to show their significant in discriminating the mentioned abnormal and normal shapes. In the result, it is found that six moment based features are significant.

Automatic leukocyte nucleus segmentation by intuitionistic fuzzy divergence based thresholding

The paper proposes a robust approach to automatic segmentation of leukocytes nucleus from microscopic blood smear images under normal as well as noisy environment by employing a new exponential intuitionistic fuzzy divergence based thresholding technique. The algorithm minimizes the divergence between the actual image and the ideally thresholded image to search for the final threshold. A new divergence formula based on exponential intuitionistic fuzzy entropy has been proposed. Further, to increase its noise handling capacity, a neighborhood-based membership function for the image pixels has been designed. The proposed scheme has been applied on 110 normal and 54 leukemia (chronic myelogenous leukemia) affected blood samples. The nucleus segmentation results have been validated by three expert hematologists. The algorithm achieves an average segmentation accuracy of 98.52% in noise-free environment. It beats the competitor algorithms in terms of several other metrics. The proposed scheme with neighborhood based membership function outperforms the competitor algorithms in terms of segmentation accuracy under noisy environment. It achieves 93.90% and 94.93% accuracies for Speckle and Gaussian noises, respectively. The average area under the ROC curves comes out to be 0.9514 in noisy conditions, which proves the robustness of the proposed algorithm.

Probabilistic prediction of malaria using morphological and textural information

The objective of this paper is to introduce a computer assisted prediction of malaria infection particularly Plasmodium vivax based on the morphological and textural information. Here erythrocytes have been segmented from light microscopic images of peripheral blood smear using marker controlled watershed followed by pre-processing. Thereafter texture and morphology of erythrocytes are extracted using geometrical and Haralick texture measure. Finally statistically significant features are fitted with multivariate regression model. This predictive model is good fitted (−2 Log Likelihood = 24.636 and Chi-Square = 820.949) and it gives 88.77 % prediction accuracy.

Plasmodium vivax segmentation using modified fuzzy divergence

This paper aims at introducing a new approach to Plasmodium vivax (P. vivax) detection from Leishman stained thin blood film. This scheme follows retrospective study design protocol where patients were selected at random in the clinic. The scheme consists of two main stages - firstly artefacts reduction, and secondly fuzzy divergence based segmentation of P. vivax infected region(s) from erythrocytes. Here, malaria parasite segmentation is done using divergence based threshold selection. Fuzzy approach is chosen to minimize ambiguity inherent in the microscopic images. Divergence algorithm is derived from Cauchy membership function to overcome the drawbacks in comparison with other well known membership functions.

Automated characterization of sub-epithelial connective tissue cells of normal oral mucosa: Bayesian approach

The objective of this paper is to develop an automated cell classification system based on Bayesian classifier followed by segmentation using color deconvolution and feature extraction for characterizing various types of sub-epithelial connective tissue (SECT) cells from histological images. In the histological sections of oral mucosa, SECT layer mainly consists of three types of cells - inflammatory, fibroblast and endothelial cells; out of which only first two play significant role pertaining to precancerous changes in oral mucosa. In order to discriminate inflammatory and fibroblast cells, a set of mathematical features viz., area, perimeter, eccentricity, compactness, Zernike moments and Fourier descriptors are extracted followed by cell segmentation using color deconvolution method. The features are statiatically analysed to show its significance in cell discrimination. Thereafter, Bayesian classifier is implemented based on the defined feature space for characterizing inflammatory and fibroblast cells in order to observe the cell distribution in healthy state. The performance of this proposed system is evaluated with 97.19% overall classification accuracy.

Development of hedge operator based fuzzy divergence measure and its application in segmentation of chronic myelogenous leukocytes from microscopic image of peripheral blood smear.

This paper introduces a hedge operator based fuzzy divergence measure and its application in segmentation of leukocytes in case of chronic myelogenous leukemia using light microscopic images of peripheral blood smears. The concept of modified discrimination measure is applied to develop the measure of divergence based on Shannon exponential entropy and Yagers measure of entropy. These two measures of divergence are compared with the existing literatures and validated by ground truth images. Finally, it is found that hedge operator based divergence measure using Yagers entropy achieves better segmentation accuracy i.e., 98.29% for normal and 98.15% for chronic myelogenous leukocytes. Furthermore, Jaccard index has been performed to compare the segmented image with ground truth ones where it is found that that the proposed scheme leads to higher Jaccard index (0.39 for normal, 0.24 for chronic myelogenous leukemia).

Entropy based divergence for leukocyte image segmentation

This work aims to develop the divergence measures based on Renyis and Yagers entropies for segmenting the leukocyte nuclei from microscopic image of peripheral blood smear. Such measure minimizes the separation between the actual and ideal thresholded image. Finally, these measures have been compared with Shannon entropy based divergence algorithm. In fact, it is observed here that Yagers measure provides better result in segmenting the leukocyte nuclei from the background of the image. The effectiveness of our proposed methods is demonstrated on blood cytopathological images of normal and chronic myelogenous leukemia (CML) samples.