An improved medical image classification model using data mining techniques

In today's world, there is a dire need for the appropriate use of technology to diagnose and treat patients by analyzing medical data, which is usually in the form of images. This need calls for an in depth research in the field of data mining and its applications for medical treatments. In this paper, an improved method to classify medical images is discussed. This method encompasses concepts related to the k-nearest neighbor (kNN) Classification algorithm and concentrates on improving the prediction ability of the algorithm using weighting techniques. This paper also uses image pre-processing techniques to select the best representative features to classify an image and to avoid the curse of dimensionality. The improved KNN algorithm is modeled using pre-processed retinal fundus images. The performance of the proposed classifier is compared with the traditional kNN classifier using metrics such as classification accuracy and area under the ROC curve.