Brain Tumor Detection Using YOLOv5 and Faster R-CNN

Abstract

Brain tumors are viewed as quite possibly the most hazardous problem in the world. Brain tumors spread quickly, and if they are not treated promptly, the patient's chances of survival are slim. Cancer cells can be benign or malignant, which is further subdivided into distinct classes such as meningioma, pituitary, and glioma. Machine-diagnosis-based methods have emerged recently and are able to identify brain cancers by utilizing magnetic resonance imaging. Two deep learning-based approaches for tumor recognition and categorization are included in our proposal, one with the YOLO (You Only Look Once) algorithm and the other using the faster R-CNN. In this case, we used YOLOv5, the fifth version of YOLO. Both methods for object detection rely on deep learning and are essentially convolutional neural networks. YOLOv5 does, however, necessarily require less computational architecture than other computing models. This paper includes a study based on the Kaggle dataset in which both models are trained across the entire dataset, and the model with the highest accuracy is used to detect brain tumors. Because YOLOv5 appears to have significantly higher precision, the dataset is trained and tested, and tumors are detected using a bounding box as well as malignancy classification using pre-trained classes. After careful calculation of the metric values, the final outcomes are shown graphically.