Detection of Brain Tumors Through the Application of Deep Learning and Machine Learning Models

Abstract

One of the most deadly conditions that can afflict both children and adults is a brain tumor. Brain tumors progress rapidly, decreasing the patient’s probability of survival if they are not effectively treated. The importance of finding brain cancers early cannot be overstated. For a variety of reasons, including the intricacy of the neuronal structure and size, as well as the form of the brain, diagnosing brain tumors can be challenging. One of the most frequent methods for detecting and localising a brain tumor is magnetic resonance imaging (MRI). Due to the complexity of brain tumors and their qualities, manual inspection might be error-prone. Therefore, early diagnosis of malignancies in the brain necessitates the development of an automated approach. To address this research challenge, this study offers a machine learning method that, given a dataset of MRI images used to classify brain tumors, can generate a labelled segmentation of those images. Four distinct categories of pictures of brain tumors are included in the dataset, including pituitary, meningioma, glioma, and no tumor. In this study, a high-pass filter was employed to demonstrate the diversity of the MRI pictures and how well they integrated with the input data. Slices were combined with a high median filter. The input MRI brain pictures were refined by smoothing them out and highlighting their borders to increase the quality of the output slices. Since the thresholding cluster matched pixels in the input MRI image, the 4-connected seed-growing technique was applied. The model is trained using supervised machine learning classification techniques, such as Support Vector Machine (SVM), K-Nearest Neighbors, Random Forest, Naive Base, and Convolutional Neural Network (CNN). The CNN technique achieved the best results.