SACMiner: A new classification method based on statistical association rules to mine medical images

Abstract

The analysis of images to decision making has become more accurate thanks to the technological progress on acquiring medical images. In this scenario, new approaches have been developed and employed in the computeraided diagnosis in order to be a second opinion to the physician. In this work, we present SACMiner, which is a new method of classification that takes advantage of statistical association rules. It works with continuous attributes and avoids introducing the bottleneck and inconsistencies in the learning model due to a discretization step, which is required in the most of the associative classification methods. Two new algorithms are employed in this method: the StARMiner* and the V-classifier. StARMiner mines association rules over continuous feature values and the V-classifier decides which class best represents a test image, based on the statistical association rules mined. The results comparing SACMiner with other traditional classifiers show that the proposed method is well-suited in the task of classifying medical images. © 2011 Springer-Verlag Berlin Heidelberg.