In this work, we propose a special cascade network for image segmentation, which is based on the U-Net networksas building blocks and the idea of the iterative refinement. The model was mainly applied to achieve higherrecognition quality for the task of finding borders of the optic disc and cup, which are relevant to the presence ofglaucoma. Compared to a single U-Net and the state-of-the-art methods for the investigated tasks, the presentedmethod outperforms others by multiple benchmarks without increasing the volume of datasets. Our experimentsinclude comparison with the best-known methods on publicly available databases DRIONS-DB, RIM-ONE v.3,DRISHTI-GS, and evaluation on a private data set collected in collaboration with University of California SanFrancisco Medical School. The analysis of the architecture details is presented. It is argued that the model canbe employed for a broad scope of image segmentation problems of similar nature.
展开▼
