Brain Tumor Detection and Segmentation Using VGG16 and Mask R-CNN with Transfer Learning

Abstract

Brain tumors are one of the leading causes of death in both adults and children in recent years, by either developing within the brain or spreading from other secondary tumors elsewhere in the body. Accurate early diagnosis of tumors can go a long way in providing better treatment options for patients. An automatic system for localization and segmentation of brain tumors can assist physicians to make a quicker and better diagnosis. In our work, we propose a deep learning model using the existing architectures of VGG16 and Mask R-CNN to detect and localize tumors in MRI-based images. With the help of transfer learning the model was capable of learning from a limited number of images to produce a test accuracy of 90% for detection and a mean average precision score of 90% for segmentation.

Keywords-Brain Tumor, Mask R-CNN, Transfer Learning, VGG16, Convolutional Neural Networks