基于深层迁移学习的DR胸片肺结核病灶检测
DOI:
作者:
作者单位:

作者简介:

通讯作者:

中图分类号:

TP391.4

基金项目:

四川省重点研发项目(18ZDYF2039)


Deep Transfer Learning based pulmonary tuberculosis lesions detection on DR films
Author:
Affiliation:

Fund Project:

  • 摘要
  • |
  • 图/表
  • |
  • 访问统计
  • |
  • 参考文献
  • |
  • 相似文献
  • |
  • 引证文献
  • |
  • 资源附件
  • |
  • 文章评论
    摘要:

    针对基于传统机器学习方法设计的DR胸片肺结核检测器存在着泛化能力不强,实际检测精度低等问题,提出了一种基于Focal Loss的深度学习检测方法Tuberculosis Neural Net(TBNN)。医学图像的特殊性,存在带标注的数据量小导致无法充分训练深层网络模型等问题。该方法利用肺炎和肺结核同为呼吸道感染疾病且在DR胸片上有相似表征的特点,基于迁移学习原理训练特征提取子网络,减少肺结核胸片样本不足对模型训练造成的影响。首先在大型的肺炎胸片数据集上训练特征提取网络,以获取DR图像中丰富的深层图像语义信息,然后使用样本较少的肺结核数据集微调网络参数,并将多层卷积的输出作为TBNN分类子网络的输入,得到基于DR胸片的肺结核病灶检测模型。实验结果表明,该方法生成的检测模型在分类精度和性能上均优于基于传统机器学习的肺结核检测器。在同等训练数据量和训练周期下,模型性能高于其他采用传统数据增强方法的深层网络肺结核检测算法,且能标识病灶区域,准度上有不低于放射科阅片医生的表现。

    Abstract:

    To find solutions to the problem that the tuberculosis detectors based on the traditional machine learning method arepoor generalization ability and low detection accuracy, this paper proposed a deep learning method based on focal loss to detect tuberculosis lesions called Tuberculosis Neural Net (TBNN), which makes full use of pneumonia and tuberculosis as respiratory diseases and has similar characteristics on chest Xray. During model training, the feature extraction subnetwork is pretrained based on transfer learning to reduce the impact of insufficient tuberculosis chest Xray samples. Firstly, the feature extraction network was trained on a large open pneumonia dataset to obtain the rich deep image semantic features. Secondly finetune the model parameters with the tuberculosis dataset and then outputs were used as inputs to the TBNN classification subnetwork to calculate tuberculosis lesion detection results. The experimental results show that TBNN is superior to the traditional machine learning methods in classification accuracy and performance. Compared with other deep learning detectors in classification accuracy, TBNN is better, and its ability to locate lesion area has reached the level of human radiologists.

    参考文献
    相似文献
    引证文献
引用本文

引用本文格式: 胡恒豪,王俊峰,方智阳,周海霞. 基于深层迁移学习的DR胸片肺结核病灶检测[J]. 四川大学学报: 自然科学版, 2020, 57: 459.

复制
分享
文章指标
  • 点击次数:
  • 下载次数:
  • HTML阅读次数:
  • 引用次数:
历史
  • 收稿日期:2019-07-25
  • 最后修改日期:2019-10-08
  • 录用日期:2019-10-28
  • 在线发布日期: 2020-05-26
  • 出版日期:
通知
自2024年3月6日起,《四川大学学报(自然科学版)》官网已迁移至新网站:https://science.scu.edu.cn/,此网站数据不再更新。
关闭