Study on texture based segmentation and clustering for prediction of medical images

Abstract

Medical images are the main source of data in healthcare applications newlineand also, they are most challenging to analyze and predict. Segmentation is the newlineessential process in image analysis for achieving characteristic extraction or newlineregion of image. While extracting the features, the grouping of similar regions newlineplays a vital role in several applications. Clustering is an unsupervised approach newlinewhich groups the images of similar regions with reduced complexity. newlineMany clustering techniques were introduced for grouping the medical images, newlinebut it still experiences certain drawbacks. Likewise, image classification is the newlinemost significant factor because it directly affects the disease prediction newlineperformance. Recently, several researches have been developed for image and newlinetexture analysis. But they fail to preserve important features of an image and newlinethus they can t able to improve the accuracy and reduce the time complexity newlinefor making effective diagnosis of disease. In order to overcome the newlineabove-mentioned issues, three different methods such as Edge-Preserved newlineDominant Valley and Discrete Tchebichef (EPDVDT) method, Kernalized newlineAverage Entropy and Density-based Spatial Clustering (KAE-DSC) method and newlineDeep Convolution Multinomial Logarithmic-based Image Classification newline(DCML-IC) method are applied with various numbers of images from database. newline