Brain Tumor Classification and Detection of Position using Deep Learning

Brain tumor classification is essential for tumor evaluation and choosing appropriate therapeutic strategies, as it is based on types of tumors. Although there are several imaging methods to identify brain tumors, Magnetic Resonance Imaging (MRI) is often more preferable because of the improved quality of the image and the usage of a non-ionizing radiation. Recently, Deep Learning (DL) has demonstrated impressive performance on segmentation and classification tasks. This was achieved by creating a Convolution Neural Network (CNN) that extracts features through the process of backpropagation with essential components. The main objectives of this work were to create a CNN for the diagnostic and segmentation purposes of a brain tumor in MRI images. The proposed classification model is developed using CNN architecture to categorize MRI images into 4 different classes such as meningioma, glioma, pituitary adenoma and non-tumor images. A U-Net architecture has been used for the segmentation model, which helps identify and separate tumor areas in MRI scans metrics. In order to test the reliability of these models, we used publicly available data, achieving impressive outcomes with a classification accuracy of 95% and a Dice Similarity Coefficient (DSC) of 93.98% for segmentation. By using the classification and the segmentation approaches, we created a web application for brain tumor categorization and localization in real time. These are promising in regards to being adapted into clinically relevant models during the diagnosis and treatment of brain tumors.