Lung cancer detection using the SOM-GRR based radial basis function neural network

Abstract

Abstract This study intends to detect lung cancer using chest X-ray. We propose the SOM-GRR based radial basis function neural network (RBFNN) model. The self-organizing maps (SOM) based deals with unsupervised learning and the global ridge regression (GRR) based deals with supervised learnings that are carried out in developing RBFNN model. The gray level co-occurrence matrix (GLCM) extraction is performed to obtain the features of chest X-ray which are used as RBFNN input variables. We consider thirteen features, namely contrast, correlation, energy, homogeneity, sum entropy, variance, inverse difference moment, sum average, sum variance, entropy, difference entropy, maximum probability, and dissimilarity. The best model is obtained by evaluating its performance in training and testing data sets. The RBFNN model yields 93 % and 88 % accuracy in training and testing data sets, respectively.