Decision support system is used in medical field to enable the physician to make quick and efficient diagnosis of diseases using information and communication technology. Quick retrieval of image sample for comparison and classification is one of the important criteria for the design of such a system. Fast decision and accuracy of the classification is always trade off. The fast decision and retrieval usually suffer from a problem of less accuracy. Here in this paper a hybrid mechanism based image retrieval cum classification is proposed which will make quick decision with guaranteed level of accuracy. The proposed system is tested on mammogram medical images for finding early detection of cancer at various stages.The computational complexity analysis result on large image data base system proves the proposed scheme can be implementable in a typical practical decision support system. © BEIESP.