BRAIN TUMOR DETECTION USING BACKPROPAGATION NEURAL NETWORKS

Abstract

A study of brain tumor detection has been done by making use of backpropagation neural networks with Gray Level Co-Occurrence Matrix (GLCM) feature extraction. CT-Scan images of the brain consist of 12 normal and 13 abnormal (tumor) brain images are analyzed. The preprocessing stage begins with cropping the image to a 256 x 256 pixels picture, then converting the colored images into grayscale images, and equalizing the histogram to improve the quality of the images. GLCM is used to calculate statistical features determined by 5 parameters i.e., contrast, correlation, energy and homogeneity for each direction. In these backpropagation neural networks, the [12 2 1] architecture is used. The correlation coefficient between the target and the output for the training data is 0.999, while the correlation coefficient for the testing data is 0.959 with an accuracy of 70%. The results of this research indicate that backpropagation neural networks can be used for the detection of brain tumors.