Detection of Cancer in Lung With K-NN Classification Using Genetic Algorithm

This paper focuses on early stage lung cancer detection. Genetic K-Nearest Neighbour (GKNN) Algorithm is proposed for the detection which is a non parametric method. This optimization algorithm allows physicians to identify the nodules present in the CT lung images in the early stage hence the lung cancer. Since the manual interpretation of the lung cancer CT images are time consuming and very critical, to overcome this difficulty the Genetic Algorithm method is combined with K-Nearest Neighbour (K-NN) algorithm which would classify the cancer images quickly and effectively. The MATLAB image processing toolbox based implementation is done on the CT lung images and the classifications of these images are carried out. The performance measures like the classification rate and the false positive rates are analyzed. In traditional K-NN algorithm, initially the distance between all the test and training samples are calculated and K-neighbours with greater distances are taken for classification. In this proposed method, by using Genetic Algorithm, K (50-100) numbers of samples are chosen for each iteration and the classification accuracy of 90% is achieved as fitness. The highest accuracy is recorded each time. (C) 2015 The Authors. Published by Elsevier Ltd.