Multi-scale Neural Networks for Retinal Blood Vessels Segmentation
This work addresses segmentation of retinal blood vessels for medical imaging applications, representing an incremental improvement over existing supervised methods.
The paper tackled retinal blood vessel segmentation by proposing a multi-scale convolutional neural network that leverages low-level and high-level features with atrous convolution, achieving significant performance improvements on DRIVE, STARE, and CHASE benchmarks in accuracy, sensitivity, specificity, AUC, and prediction speed.
Existing supervised approaches didn't make use of the low-level features which are actually effective to this task. And another deficiency is that they didn't consider the relation between pixels, which means effective features are not extracted. In this paper, we proposed a novel convolutional neural network which make sufficient use of low-level features together with high-level features and involves atrous convolution to get multi-scale features which should be considered as effective features. Our model is tested on three standard benchmarks - DRIVE, STARE, and CHASE databases. The results presents that our model significantly outperforms existing approaches in terms of accuracy, sensitivity, specificity, the area under the ROC curve and the highest prediction speed. Our work provides evidence of the power of wide and deep neural networks in retinal blood vessels segmentation task which could be applied on other medical images tasks.