首页> 外文会议>International conference on neural information processing >Combining CNN and MIL to Assist Hotspot Segmentation in Bone Scintigraphy
【24h】

Combining CNN and MIL to Assist Hotspot Segmentation in Bone Scintigraphy

机译:结合CNN和MIL协助骨骼闪烁成像中的热点分割

获取原文

摘要

Bone scintigraphy is widely used to diagnose tumor metas-tases. It is of great importance to accurately locate and segment hotspots from bone scintigraphy. Previous computer-aided diagnosis methods mainly focus on locating abnormalities instead of accurately segmenting them. In this paper, we propose a new framework that accomplish the two tasks at the same time. We first use sparse autoencoder and convolution neural network (CNN) to train an image-level classifier that label input image as normal or suspected. For suspected images, multiple instance learning (MIL) is applied to train a patch-level classifier. Then we use this classifier to produce a probability map of hotspots. Finally, level set segmentation is performed with the probability map as initial condition. The experimental results demonstrate that our method is more accurate and robust than other methods.
机译:骨闪烁显像术被广泛用于诊断肿瘤转移。准确地定位和分割来自骨闪烁显像仪的热点非常重要。以前的计算机辅助诊断方法主要关注于定位异常,而不是准确地对其进行分段。在本文中,我们提出了一个新的框架,该框架可以同时完成两个任务。我们首先使用稀疏自动编码器和卷积神经网络(CNN)来训练将输入图像标记为正常或可疑的图像级分类器。对于可疑图像,应用多实例学习(MIL)来训练补丁程序级别的分类器。然后,我们使用此分类器生成热点的概率图。最后,以概率图作为初始条件执行水平集分割。实验结果表明,我们的方法比其他方法更准确,更可靠。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号