Brain Tumor Detection Using Convolutional Neural Network

Abstract

A brain tumor (BTR) is the development of aberrant and uncontrolled cells in the brain. The detection of a BTR in its early stages is essential in the treatment of its sufferers. There are various ways to diagnose a BTR but Imaging is one of the accurate ways to find the critical one. There are various imaging tests available like Magnetic Resonance Imaging (MRI), Computerised Tomography (CT) scan, and Positron Emission Tomography (PET). MRI is preferable among all because it is highly adept at capturing images that help doctors determine if there are abnormal tissues within the body. Detecting BTR by just looking into MRI images is prone to human errors and the patient may reach the end stage of the disease. Therefore, the main objective of this research is to create a Convolutional Neural Network (CNN) that can detect and classify whether a patient has a BTR or not. In the proposed method, ‘Leaky ReLU’ activation function with convolution 2D layer (Conv2D + Leaky ReLU) combine and compares the model accuracy with a pre-implemented CNN model i.e., (Conv2D + ReLU) layers combinations. The proposed model achieved 78.57% validation accuracy, which is higher than the normal pre-implemented CNN model. However, the training accuracy score of both the model is 99.20%.