Comparative study of clustering medical images

Due to the fast growing of the images data repositories, there is a big challenge to organize these repositories and make them easy to search and mine to get knowledge. Image clustering goal is to link each image in an image database with a class label, so similar images are grouped and have the same class label and which are different from other images in a database. Image clustering organize image repositories to groups or clusters and this is a basic step in many applications as content based image retrieval (CBIR), image classification, powerful search engines, browsing images and segmentation of images. In this paper, we proposed two new methods for clustering medical images. The two proposed methods are implemented, tested on medical images data set and the results obtained from the two proposed methods are compared. In the first proposed method, GLCM and Haralick's statistical measures are used for extracting texture features from the images and k-means clustering is applied to cluster the extracted feature vectors. In the second proposed method, 2D wavelet transform is used to extract features from the images, feature selection is applied and finally, K-means clustering algorithm is applied to cluster feature vectors.