Brain tumors are highly fatal and diagnosis can often be time-consuming. Timely attention and proper consultation is required to successfully detect the tumor. However, the detection of brain tumors is a very tedious task. Brain tumors are of numerous types which pose a challenging task of detection and classification as the tumors can be ill-defined with soft tissues. Magnetic Resonance Imaging (MRI), Computerized Tomography (CT) scans, and Ultrasound images are generally used to get the brain images. Brain tumors are common, and because of the associated risk of malignancy and hyperfunction, these tumors have to be examined thoroughly. To automate the process of detection and further segmentation, a robust system is required which can be used to produce accurate results. In this work, this observation is taken into consideration and a technique was proposed that will bridge the gap between diagnosing and detecting tumors. This process will further provide a valuable second opinion to medical professionals. The methodology involves brain tumor detection using transfer learning using Residual Networks (ResNet). Further, segmentation is done to identify the area of the tumor in the MRI scan. This work demonstrates Deep Learning’s potential in processing and extracting information from MRI images to provide a non-invasive tool for automated tumor detection and segmentation for clinical applications.

Deep Learning Brain Tumor Segmentation Brain Tumor Detection MRI Neural Networks.