Abstract:
Retinal vascular image segmentation plays an important role in the automatic diagnosis of diseases, and is a key step in early diagnosis and surgical planning. Thus, accurate segmentation of the retinal vascular tree has become a prerequisite for computer-aided diagnosis. With the application of convolutional neural networks in medical image segmentation, some Network with optimized segmentation performance have been gradually proposed. However, these methods ignore the attention of context and multi-view, making it difficult to separate microvascular branches. To solve this problem, a multi-view context attention network architecture is proposed. Firstly, the network adds a new multi-view attention module, which can effectively focus on the valid information while expanding the receptive field, to prevent the fracture of microvascular segmentation. Secondly, the network integrates attention gate, it combines low-level features with high-level features, produced more representative new features, and further improve the performance of microvascular segmentation. It is evaluated on two public fundus datasets DRIVE and CHASE DB1, the experimental results show that the algorithm is superior to the state-of-the-art methods in accuracy, sensitivity, intersection-over-union and AUC (area under ROC curve).