Enhanced brain tumor detection using fractional wavelet transform and artificial neural network

Abstract

Abstract The timely identification of brain tumor in individuals is an important but difficult task. Hence, researchers are coming up with new techniques for early brain tumor detection to increase the possibility of survival of patients which are affected by it. This book chapter proposes an enhanced brain tumor detection method based on fractional wavelet transform (FrDWT) to obtain the features, principal component analysis to lower the dimensionality of features, and artificial neural network for classification. In this method, two-dimensional FrDWT with third level decomposition and fractional angle α is applied on the brain MR images on approximation coefficients. Furthermore, principal component analysis (PCA) is implemented. The classification is performed with the help of artificial neural network (ANN). The FrDWT algorithm is evaluated using two datasets. The MR images for dataset Db1 is obtained from the brain atlas Website of Harvard Medical School and for dataset Db2 the images are taken from BraTS2015 and e-health laboratory. As it is a binary classification problem, the labels “normal” and “abnormal” are used to categorize the entire training dataset. The experimental observations of fractional wavelet transform with the values of fractional angle “α” in the range of 0.1–1 are compared with the conventional DWT. The algorithm is evaluated using different parameters including sensitivity, specificity, accuracy, precision, and F1 score.