Implementation of medical decision support system for classification of brain tumor using image mining techniques

Abstract

The development of Medical Decision Support System (MDSS) is becoming more important to support medical practitioners for better health care in the recent years. It has been witnessed that millions of medical images have been generated in medical care centers worldwide and are provided for medical diagnosis. Analysis of large data and manual interpretation become very challenging, even for the experienced and trained sonologist. Hence, physicians prefer computer based techniques for diagnosing the patients. This has initiated the research to create a vast database and implement various techniques for medical image analysis. The Computerized Tomography (CT) has been found to be the most reliable method for early detection of tumors. The present work consists of four phases: training phase, mining phase, test phase, and classification phase. In the training phase, acquired CT scan brain images are processed using the histogram equalization to improve the image quality. In the association rule mining phase, the association rules have been constructed from the stored feature vectors. This transaction representation is submitted to the proposed Mining Association Rule in Image database (MARI) and ImageApriori algorithm for association rule mining. In order to validate the obtained results, the algorithmic approach has been compared with the well known classifiers, a naive Bayesian classifier and associative classifier. The experimental results have shown that the proposed methods achieve high sensitivity (up to 97%), specificity (up to 91%) and accuracy (up to 98.5%). The average training and average testing time have been reduced to 4.39 sec and 1.01sec of CPU time respectively. Thus it is concluded that the brain images can be processed and tumor can be classified using image mining techniques for better decision making. newline newline newline