DDNet: A Novel Network for Cerebral Artery Segmentation from MRA Images

Brain vascular segmentation is a very important step in the diagnosis and treatment of brain diseases. Since manual segmentation of the vascular system of the brain is laborious and has problems with accuracy, automatic or semi-automatic cerebrovascular segmentation is of great significance for the study of brain diseases. Compared to hand-defined features, deep learning methods are widely used in medical image processing due to their superior feature expression capabilities. Many of the existing deep learning-based blood vessel segmentation methods do not consider multi-scale information of blood vessels, and the extracted context information has a small receptive field. Besides, the network parameters are large and difficult to train. In this paper, we proposed a network named Dense-Dilated Neural Network (DDNet) for the segmentation of cerebral artery in TOF-MRA images. The backbone of DDNet is U-Net, consisting of an encoding path, a decoding path and the skip connections between the two paths. The network is able to obtain a multi-scale representation of the brain arterial network, and has a lager respective field. In addition, the final brain arterial network segmentation results can be obtained with less computational effort. Experiments show that the network proposed in this paper is superior to UNet, Vnet and Uception.