Advancing Brain Tumor Classification: Exploring Two Deep Learning Architectures for Improved Accuracy

A mass of abnormal cells that form inside or outside the brain is called a brain tumor. Adults are at high risk of developing brain tumors, which can cause serious organ dysfunction and even death. Detecting tumors manually is a tedious and difficult method that might yield erroneous findings. As a result, these tumors must be meticulously classified to offer a complete medical diagnosis and design an appropriate treatment plan. Estimating the patient's chances of survival is difficult since tumors are uncommon and can vary greatly in size, location, and history. In order to address these issues, the use of two different deep learning frameworks for multi-class brain tumor classification utilizing Magnetic Resonance Imaging (MRI) data was examined in this study. Significant evaluation metrics, including F1 score, recall, accuracy, and precision, were applied to these models. Both models demonstrated significant improvements over prior brain tumor classification studies, illustrating that deep learning algorithms may be used in the future to accurately diagnose brain tumors and enable medical personnel to make well-informed judgments regarding patients' treatment courses. This study proposes two classification algorithms: ResNet50, that obtained a success rate of 99.39%, and EfficientNetB0, obtained accuracy rate of 99.75%.