Multimodal Brain Tumor Segmentation Using a 3D ResUNet in BraTS 2021

Abstract

In this paper, we propose a multimodal brain tumor segmentation using a 3D ResUNet deep neural network architecture. Deep neural network has been applying in many domains, including computer vision, natural language processing, etc. It has also been used for semantic segmentation in medical imaging segmentation, including brain tumor segmentation. In this work, we utilize a 3D ResUNet to segment tumors in brain magnetic resonance image (MRI). Multimodal MRI is prevailing in brain tumor analysis due to providing rich tumor information. We apply the proposed method to the Multimodal Brain Tumor Segmentation Challenge (BraTS) 2021 validation dataset for tumor segmentation. The online evaluation of brain tumor segmentation using the proposed method offers the dice score coefficient (DSC) of 0.8196, 0.9195, and 0.8503 for enhancing tumor (ET), whole tumor (WT), and tumor core (TC), respectively.