高智勇,万昕.基于深度学习的肺结节识别[J].中南民族大学学报自然科学版,2019,(3):393-396
基于深度学习的肺结节识别
Recognition of pulmonary nodule based on deep learning
  
DOI:10.12130/znmdzk.20190314
中文关键词: 肺结节  深度学习  小样本  迁移训练
英文关键词: lung nodule  deep learning  small sample  transfer training
基金项目:湖北省自然科学基金资助项目(2014CFB922)
作者单位
高智勇,万昕 中南民族大学 生物医学工程学院武汉 430074 
摘要点击次数: 58
全文下载次数: 62
中文摘要:
      针对肺结节识别问题,采用分支结构改进直接型VGG16的深度学习网络结构,并使用迁移训练,减少了网络结构复杂度.结果提高了学习效率,降低了方法对硬件资源的要求,增加了其适用性.通过训练多种网络模型,对比模型特点,调整改进的网络结构实现了对小样本的肺结节识别.在LIDC-IDRI数据集上的结果表明:该方法在保持结构简单、降低硬件资源需求的同时,取得较好的肺结节识别效果.
英文摘要:
      To identify small pulmonary nodules, a modified branch structure was used to improve the deep learning network structure of the direct VGG16, and migration training was used to reduce the complexity of the network structure. The results showed that the learning efficiency was improved, the requirement for hardware resources was reduced, and the scope of application was increased. By training a variety of network models and comparing the characteristics of the model, the recognition of the small pulmonary nodules was realized. The results on the LIDC-IDRI data set showed that this method not only kept the structure simple and reduced the requirement of hardware resources, but also achieved better recognition of lung nodules.
查看全文   查看/发表评论  下载PDF阅读器
关闭