• 主题
高级搜索  >
历史搜索 清空历史
首页> 外文会议>Society of Photo-Optical Instrumentation Engineers;SPIE Medical Imaging Conference >Cancer Detection in Mass Spectrometry Imaging data by Dilated Convolutional Neural Networks
【24h】

Cancer Detection in Mass Spectrometry Imaging data by Dilated Convolutional Neural Networks

机译：扩散卷积神经网络在质谱成像数据中检测癌症

获取原文
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

Imaging mass spectrometry (IMS) is a novel molecular imaging technique to investigate how molecules are dis-tributed between tumors and within tumor region in order to shed light into tumor biology or nd potentialbiomarkers. Convolutional neural networks (CNNs) have proven to be very potent classifiers often outperformingother machine learning algorithms, especially in computational pathology. To overcome the challenge of com-plexity and high-dimensionality of the IMS data, the proposed CNNs are either very deep or use large kernels,which results in large amount of parameters and therefore a high computational complexity. An alternative isdown-sampling the data, which inherently leads to a loss of information. In this paper, we propose using dilatedCNNs as a possible solution to this challenge, since it allows for an increase of the receptive eld size, neitherby increasing the network parameters nor by decreasing the input signal resolution. Since the mass signature ofcancer biomarkers are distributed over the whole mass spectrum, both locally- and globally-distributed patternsneed to be captured to correctly classify the spectrum. By experiment, we show that employing dilated convo-lutions in the architecture of a CNN leads to a higher performance in tumor classication. Our proposed modeloutperforms the state-of-the-art for tumor classication in both clinical lung and bladder datasets by 1-3%.
机译：成像质谱（IMS）是一种研究分子是如何解释的新型分子成像技术 在肿瘤和肿瘤区域内致敬，以落入肿瘤生物学或ND潜力 生物标志物。卷积神经网络（CNNS）已被证明是非常有效的分类器经常表现优于表现 其他机器学习算法，尤其是在计算病理学中。克服COM-的挑战 IMS数据的Ploxity和高度，所提出的CNN是非常深的或使用大核， 这导致大量参数，因此是高计算复杂性。另一种选择是 向下采样数据，其固有地导致信息丢失。在本文中，我们建议使用扩张 CNNS作为这种挑战的可能解决方案，因为它允许增加接收器尺寸，既不是 通过增加网络参数，也可以通过降低输入信号分辨率来增加。自大众签名以来 癌症生物标志物分布在整个质谱上，包括局部和全球分布的图案 需要捕获以正确分类频谱。通过实验，我们表明雇用扩张的沟通 CNN架构中的原因导致肿瘤分类的表现较高。我们提出的模型 在临床肺和膀胱数据集中的肿瘤分类优于1-3％的最先进的肿瘤分类。

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