Recognizing Brain Tumor Using Convolutional Neural Networks

Abstract

A brain tumor is a collection of abdominal cells in the brain that, if left untreated, can cause permanent brain damage and even death. In order to stop the progression of this life-threatening malignant tumor, early detection and treatment are essential. Conventionally, brain specialists find any tumor abnormalities by physically studying medical images such as MRI images. This is how they detect any tumors. However, due to the cost of educating trained staff as well as other normal human limits such as lack of speed and exhaustion, a delayed or mistaken identification of the tumor may result, which in turn may lead to a delay in the initiation of treatment. In recent years, machine learning has been utilized in the diagnosis of human illnesses such as fetal malformations and abnormalities in red blood cells. Convolutional neural networks are the focus of this research as a potential way to detect brain tumors from MRI scans. Because of the limited amount of time available, the focus of this research is on the detection and identification of tumors as either benign or malignant. Benign and malignant tumors are the most prevalent types of cancerous tumors that can be discovered and rectified at an early stage. The primary purpose of this research is to construct a deep learning model that is capable of classifying MRI images as either benign or malignant based on their examination. In total, 3920 different MRI image samples were utilized for the project. The image dataset was divided into a training set and a testing set. A hyper-parameter tuning for the model was performed with the following settings for model training: 0.001 learning rate, 36 epochs, 80 steps per epoch, and an 80-step batch size. The model achieved an accuracy of 98.6 percent, with a precision of 99 percent and a recall of 98 percent, respectively