[1]曹帅,严加勇△,崔崤峣,等.基于Hessian矩阵与多视图卷积神经网络的纵隔淋巴结自动检测方法*[J].生物医学工程研究,2021,02:131-137.
CAO Shuai,YAN Jiayong,CUI Yaoyao,et al.Automatic detection method for mediastinal lymph nodes based on Hessian matrix and multi-view convolution neural network[J].Journal of Biomedical Engineering Research,2021,02:131-137.
点击复制
基于Hessian矩阵与多视图卷积神经网络的纵隔淋巴结自动检测方法*(PDF)
《生物医学工程研究》[ISSN:1006-6977/CN:61-1281/TN]
- 期数:
-
2021年02期
- 页码:
-
131-137
- 栏目:
-
- 出版日期:
-
2021-06-25
文章信息/Info
- Title:
-
Automatic detection method for mediastinal lymph nodes based on Hessian matrix and multi-view convolution neural network
- 文章编号:
-
1672-6278 (2021)02-0131-07
- 作者:
-
曹帅1; 严加勇2; 3; 4△; 崔崤峣4; 于振坤5
-
1.上海理工大学医疗器械与食品学院,上海 200093;2.上海健康医学院医疗器械学院,上海 201318;3.上海健康医学院附属周浦医院,上海201318;4.中科院苏州生物医学工程技术研究所,苏州 215163;5.南京同仁医院,南京211102
- Author(s):
-
CAO Shuai1; YAN Jiayong2; 3; 4; CUI Yaoyao4; YU Zhenkun5
-
1. School of Medical Instrument and Food Engineering,University of Shanghai for Science and Technology, Shanghai 200093,China; 2.School of Medical Instrument ,Shanghai University of Medicine & Health Science, Shanghai 201318;3.Affiliated Zhoupu Hospital, Shanghai University of Medicine &Health Science, Shanghai 201318;4. Institute of Biomedical Engineering and Technology, Chinese Academy of Sciences, Suzhou 215163,China;5. Tongren Hospital of Nanjing, Nanjing 211102,China
-
- 关键词:
-
纵隔淋巴结检测; 感兴趣区域; Hessian矩阵; 多尺度增强; 卷积神经网络; CT图像
- Keywords:
-
Mediastinal lymph node detection; Region of interest; Hessian matrix; Multiscale enhancement; Convolution neural network; CT images
- 分类号:
-
R318
- DOI:
-
10.19529/j.cnki.1672-6278.2021.02.05
- 文献标识码:
-
A
- 摘要:
-
淋巴结检测和分析对于肿瘤的疗效评估、分期具有重要意义。本研究提出一种基于Hessian矩阵与多视图卷积网络纵隔淋巴结自动检测方法。该方法首先确定纵隔部位淋巴结可能存在的区域; 然后,基于淋巴结形态特征,构造多尺度增强滤波器,提取候选淋巴结; 最后结合CT图像的冠状面、横断面和矢状面信息,设计多视图卷积网络对候选淋巴结进行分类。对90组患者的纵隔部位进行测试,平均每个患者9个假阳性淋巴结的条件下,灵敏度为90.32%。该方法对不同大小的淋巴结具有较高的检出率。
- Abstract:
-
Lymph node detection and analysis are important for the assessment of cancer efficacy and staging. We proposed an automatic lymph node detection method based on Hessian matrix and multi-view convolutional network for mediastinal lymph nodes. Firstly, lymph nodes in the mediastinal region were identified. Then a multi-scale enhancement filter was constructed to obtain the candidate lymph nodes based on the morphological characteristics of lymph nodes.Finally, a multi-view convolution network was designed to classify the candidate lymph nodes by combing the coronal, transverse and sagittal information of CT images.The mediastinum of 90 groups patients were tested, the sensitivity was 90.32% under the condition of 9 false positive lymph nodes per patient.This method has high detection rate for different size lymph nodes.
备注/Memo
- 备注/Memo:
-
(收稿日期:2020-09-25)上海市浦东新区科技发展基金民生科研专项医疗卫生项目(PKJ2017-Y41);江苏省省级重点研发专项资金资助项目(BE2017601)。△通信作者Email: yanjy@sumhs.edu.cn
更新日期/Last Update:
2021-07-21