Texture Classification of Biased Cytoplasmic Tissue Sample from Histopathological Imagery by Gabor

Abstract

Recent research with H&E stained imagery led to rapid progress towards quantifying the perceptive issues, while prognostic, due to subjective variability among readers. This  variability leads to distinguish prognosis report and variability in treatment as well. So in this study we present thoughtful analysis of texture heterogeneity on Follicular Lymphoma and Neuroblastoma tissue images for the purpose of identifying regions of interest in tissue for morphological behavior. We are introducing a classification approach for determining the texture feature i.e. described by parameters like nuclei, cytoplasm, extracellular material and red blood cells and the subsequent classification of histopathological digital image. Basic idea behind this research is to distinguish among nuclei, cytoplasm,
extracellular material and red blood cells from H&E stained input image so that doctors (radiologist) can provide better judgment during the prognosis of histopathological image that sometimes wrongly concluded (even though educated). In this memorandum we proposed a noble algorithm in which we convolve our H&E stained pathological images with 12 different orientation masks (masks obtained from Gabor application), resulting in an outputs of 12 different representations (corresponding to 12 different orientations masks) of our H&E stained input image i.e. the information included in the 12 representations coming from the application of Gaussian filter is summarized in twelve images that correspond to each of the orientations used in the filters. We then combine these 12 images
into one textured image represented as a 3-dimensional representation of input image. Experimental results on FL & NB demonstrate that the proposed approach outperforms the gray level based texture analysis.