Multi-branch Context Awareness Network for Prostate MRI Segmentation

Abstract Prostate image segmentation is a precondition for the diagnosis of prostate diseases and subsequent treatment. However, blurry organ edges low contrast make accurate difficult. In order to overcome difficulties, this paper proposes an innovative Multi-branch Context Awareness Network (MBCA-Net) MRI segmentation. MBCA-Net uses 3D UNet as backbone network with encoder-decoder framework. To improve feature extraction capacity network, multi-branch residual module different convolution kernels proposed. better use extracted features reduce semantic ambiguity, context awareness used fuse low-level in encoder high-level decoder full process manner. stability model during training, mixed loss function adopted. Experiments on public dataset PROMISE12 show that our model’s Dice coefficient 90.19%. Compared other advanced techniques, proposed has results.