Automated Analysis of Structural Alterations in Microscopic Cellular Images for Identification of Cytotoxicity Using Texture Feature Descriptors

Cytotoxicity causes alteration in cellular structures due to the action of certain chemicals. Differentiation of fluorescently labelled nuclei of healthy and cytotoxic microscopic cell images is a challenging task due to the subtle differences in cell morphology. The present work attempts to distinguish between healthy and toxic cell nuclei images using texture feature descriptors and Support Vector Machine classifier. For this, fluorescence microscopic images of Mouse cardiac muscle cells are considered from a public dataset. Texture features from these images are computed using Grey Level Run Length Matrix. Statistical analysis is carried out to select significant features. The results indicate that all the extracted features exhibit statistical significance in differentiating healthy and toxic images. The Run Percentage feature illustrates a maximum percentage difference for healthy and toxic images. Further, a classification accuracy of 93.8% and sensitivity of 95.5% is achieved. Experimental results demonstrate the ability of texture features in characterizing healthy and toxic microscopic cell nuclei images.