This website requires JavaScript.

OMSN and FAROS: OCTA Microstructure Segmentation Network and Fully Annotated Retinal OCTA Segmentation Dataset

Peng XiaoXiaodong HuKe Ma ...+3 Jin Yuan
Dec 2022
摘要
The lack of efficient segmentation methods and fully-labeled datasets limitsthe comprehensive assessment of optical coherence tomography angiography (OCTA)microstructures like retinal vessel network (RVN) and foveal avascular zone(FAZ), which are of great value in ophthalmic and systematic diseasesevaluation. Here, we introduce an innovative OCTA microstructure segmentationnetwork (OMSN) by combining an encoder-decoder-based architecture withmulti-scale skip connections and the split-attention-based residual networkResNeSt, paying specific attention to OCTA microstructural features whilefacilitating better model convergence and feature representations. The proposedOMSN achieves excellent single/multi-task performances for RVN or/and FAZsegmentation. Especially, the evaluation metrics on multi-task modelsoutperform single-task models on the same dataset. On this basis, a fullyannotated retinal OCTA segmentation (FAROS) dataset is constructedsemi-automatically, filling the vacancy of a pixel-level fully-labeled OCTAdataset. OMSN multi-task segmentation model retrained with FAROS furthercertifies its outstanding accuracy for simultaneous RVN and FAZ segmentation.
展开全部
图表提取

暂无人提供速读十问回答

论文十问由沈向洋博士提出,鼓励大家带着这十个问题去阅读论文,用有用的信息构建认知模型。写出自己的十问回答,还有机会在当前页面展示哦。

Q1论文试图解决什么问题?
Q2这是否是一个新的问题?
Q3这篇文章要验证一个什么科学假设?
0
被引用
笔记
问答