Enhanced Brain Tumor Detection and Classification in MRI: A CNN and Morphological Feature Approach

Abstract

The number of diseases associated with brain tumors has increased significantly over the past few years, making them the tenth most common cancer affecting children and adults. Because brain tumors vary in size, mass, and location, brain diagnosis and classification are the most important and time-consuming tasks in diagnosis. Magnetic Resonance Imaging (MRI) is widely used to diagnose tumors and various soft tissue abnormalities in many diseases. Examining the size and location of brain tumors plays an important role in brain diagnosis. In this paper, a deep learning method is used to segment and classify brain tumors via MRI. Firstly, the image is preprocessed using image enhancement using a median and bilateral filter. Binary threshold is then used for segmentation. Morphological functions are used for feature extraction. Finally, Convolutional Neural Network (CNN) is used to predict whether the brain MRI was normal or abnormal. Multiple brain MRIs are used, including tumor and healthy brain images, to train the model and the model achieves 96% accuracy.