Brain Tumor Detection Using Wavelet Transform and Convolution Neural Network in Magnetic Resonance Imaging

The brain tumor is the growth of an abnormal mass in the brain which may be malignant or benign. Misdiagnosis of different types of brain tumors reduces the patients' chances of survival. In the last years, efforts have been made to automate the present process for clinical and research reasons. For this purpose, machine learning methods were highly considered for a long time. In this paper, we presented a method for classification of brain tumors from magnetic resonance images using a convolutional neural network and two-level discrete wavelet transform algorithm. The BRATS 2013 dataset contains FLAIR, T1, T1c and T2 images of 20 patients with high-grade glioma, and ten patients with low-grade glioma were used in this investigation. This study aimed to classify the data into two classes of high-grade glioma and low-grade glioma by convolution neural network using a two-level wavelet transform algorithm on images. The Image wavelet coefficients contain more information than the original images, thus increasing the network performance. All the experiments were performed on the main dataset and two synthetic datasets with 60 and 300 data. The highest classification accuracy is obtained for T1 images in all three sets of 30, 60 and 300 data that were 86.25%, 94.58%, and 95.86%, respectively.