基于神经网络乳腺解剖层分割试验设计
首发时间:2018-11-23
摘要:本文采用了3D标注技术,提出三维的神经网络用于乳腺解剖层分割法,该方法巧妙地运用3DUnet技术实现端到端的乳腺解剖层分割。通过对100套三维超声数据模拟自动分割的统计试验, 试验结果显示:新方法比原来最好方法的效果提升了10%, 同时分割结果更加平滑和稳定,避免了局部剧烈的变动,降低了医生的工作负担,试验结果证明了新方法具有重要的临床意义。
For information in English, please click here
Layer Parsing of Breast Anatomy based on neural network
Abstract:We write 3D markup tool and propose 3D neural network to be used in layer parsing problem, which utilizes 3D Unet technique to realize it end to end. We prove that our method is 10% better than state of the art method and our prediction result is more smooth and stable which is able to avoid drastic changes.It is useful to reduce doctors\' load so that it have a significant influence in clinical applications.
Keywords: Breast cance Breast layer segmentation Unet
基金:
引用
No.****
动态公开评议
共计0人参与
勘误表
基于神经网络乳腺解剖层分割试验设计
评论
全部评论0/1000