人工智能培训

搜索

人工智能论文:连接敏感注意U-NET用于准确的视网膜血管分割(Connection Sensitive Attention U-NET for Accurate

[复制链接]
dabiao2008 发表于 2019-3-15 13:05:32 | 显示全部楼层 |阅读模式
dabiao2008 2019-3-15 13:05:32 371 0 显示全部楼层
人工智能论文:连接敏感注意U-NET用于准确的视网膜血管分割(Connection Sensitive Attention U-NET for Accurate Retinal Vessel  Segmentation)我们开发了一种连接敏感的注意力U-Net(CSAU),用于准确的视网膜血管分割。该方法通过四个关键改进改进了近期关注的U-Net语义分割:(1)连接敏感性,对结构属性进行建模,提高像素分割的准确性; (2)注意门具有新颖的神经网络结构,并与DOWN-Link相结合,有效地学习了对细血管的更好的注意力; (3)连接敏感损失和注意门的集成通过在输出之前额外连接特征权重来提高对详细容器的准确性; (4)连接灵敏度的指标,以反映边界和血管上的分割性能。我们的方法可以有效地改善现有的血管分割方法,这些方法存在异常,分叉和微血管的困难。这种连接敏感性损失与提出的注意力U-Net紧密结合,以准确地(i)分割视网膜血管,以及(ii)通过对结构特性建模来保留薄血管的连通性。我们的方法在DRIVE,STARE和HRF数据集上取得领先地位在最先进的方法之中。
We develop a connection sensitive attention U-Net(CSAU) for accurate retinalvessel segmentation.This method improves the recent attention U-Net forsemantic segmentation with four key improvements: (1) connection sensitive lossthat models the structure properties to improve the accuracy of pixel-wisesegmentation;(2) attention gate with novel neural network structure andconcatenating DOWN-Link to effectively learn better attention weights on finevessels;(3) integration of connection sensitive loss and attention gate tofurther improve the accuracy on detailed vessels by additionally concatenatingattention weights to features before output;(4) metrics of connectionsensitive accuracy to reflect the segmentation performance on boundaries andthin vessels.Our method can effectively improve state-of-the-art vessel segmentationmethods that suffer from difficulties in presence of abnormalities, bifurcationand microvascular.This connection sensitive loss tightly integrates with theproposed attention U-Net to accurately (i) segment retinal vessels, and (ii)reserve the connectivity of thin vessels by modeling the structural properties.Our method achieves the leading position on DRIVE, STARE and HRF datasetsamongthe state-of-the-art methods.人工智能论文:连接敏感注意U-NET用于准确的视网膜血管分割(Connection Sensitive Attention U-NET for Accurate Retinal Vessel  Segmentation) TR133Hf6Q3F50JiZ.jpg
URL地址:https://arxiv.org/abs/1903.05558     ----pdf下载地址:https://arxiv.org/pdf/1903.05558    ----人工智能论文:连接敏感注意U-NET用于准确的视网膜血管分割(Connection Sensitive Attention U-NET for Accurate Retinal Vessel  Segmentation)
回复

使用道具 举报

您需要登录后才可以回帖 登录 | 立即注册

本版积分规则 返回列表 发新帖

dabiao2008当前离线
新手上路

查看:371 | 回复:0

快速回复 返回顶部 返回列表