Squeeze Excitation Embedded Attention U-Net for Brain Tumor Segmentation

Deep learning-based techniques have gained significance over the past few years in the field of medicine. They are used in various applications such as classification and segmentation of medical images. Existing architectures such as U-Net, Attention U-Net, and Attention Residual U-Net are the state-of-the-art architectures for brain tumor segmentation. However, these architectures do not address the issue of how to extract the features in channel level. In this work, we propose a new architecture called Squeeze Excitation Embedded Attention U-Net (SEEA-U-Net) which incorporates squeeze excitation network in Attention U-Net for better segmentation results. SEEA-U-Net extracts information at both spatial and channel levels. It outperforms the state-of-the-art architectures in terms of binary focal loss and Jaccard coefficient, particularly when the number of epochs is limited.