Detection Of Brain Tumor In MRI Images Using Deep Learning Techniques

Abstract

Most of the deaths in the world happen due to cancer. It is a disease in which the cells of our body organs or tissues grow in an undisciplined manner which in turn can harm our normal cells and tissues in our body. These cells very smartly trick the immune system so that the cancerous cells are kept alive and are not destroyed. In the human body, tumors can be classified into three types: cancerous, non-cancerous, and pre-cancerous. Timely identification of the cancer can be helpful in many ways. As it improves a patient s chances of survival. The most valuable, uncomplicated technique used is MRI scans for predicting tumor. It is a tough task and has chances of human error. So to be more accurate with our predictions, moved to use computerised techniques to ease the work. One of the malignant cancers is a brain tumor. Brain tumor can affect people of all ages. There are many types of cancer based on factors such as shape, texture, size, etc. The detection of cancer is done by MRI scan which is Magnetic Resonance Imaging. This method gives a clear picture of the brain in 2D or 3D format. This method is popular and precise for detecting cancer. Just looking at the images and then trying to predict the type of tumor is a tough job, and if done manually, there are chances of human error. The use of Machine Learning and Deep Learning techniques can improve the accuracy of detecting cancer and, hence, save many lives. Different models, like CNN, Alex Net, etc. provide different accuracy and precision when used with different machine learning models, such as SVM, k-NN, etc. The focus of this research is the development of an automated brain tumor classification system using magnetic resonance imaging (MRI) scans, leveraging a deep learning model. Recent advancements in technology have enabled the utilisation of these methods for detecting tumors present in the brain. In this research, developed fine-tuned MobileNet base model to get a faster result. The precision and accuracy of the model were enhanced by restructurin