Optimized association rules for MRI brain tumor classification

Concept of data mining for discovering frequent image patterns present in MRI brain Tumor image data base is presented. In the proposed work, an image mining method using optimized association rules for effective classification of brain Tumor MRI images is presented. It consists of finding frequent image patterns and classifies tumor images using optimized association rules. It consists of finding the features and optimizes the number of features by selecting most discriminating features among them. These features are further discretized and generate transaction representation of input images. This is given as input to Apriori algorithm for generating association rules. We use multi objective genetic (MOGA) algorithm to optimize and get strong, highly correlated association rules. Result show that image mining is feasible and gives strong association rules, reduces computational complexity of mining process. Optimized association rules are used for effective classification of MRI images. Proposed algorithm gives better accuracy around 90%.