Brain Tumor Segmentation Using 3D Convolutional Neural Network

Brain tumors segmentation is one of the most crucial procedures in the diagnosis of brain tumors because it is of great significance for the analysis and visualization of brain structures that can guide the surgery. With the development of natural scene segmentation model FCN, the most representative model U-net has been developed. An increasing number of people are trying to improve the encoder-decoder architecture to achieve better performance currently. In this paper, we focus on the improvement of the encoder-decoder network and the analysis of 3D medical images. We propose an additional path to enhance the encoder part and two separate up-sampling paths for the decoder part of the model. The proposed approach was trained and evaluated on BraTS 2019 dataset.