[1]冯翔△,康文清,吴瀚,等.基于深度特征融合的肺炎影像识别研究*[J].生物医学工程研究,2021,01:28-32.
FENG Xiang,KANG Wenqing,WU Han,et al.Research on pneumonia image recognition based on deep feature fusion[J].Journal of Biomedical Engineering Research,2021,01:28-32.
点击复制
《生物医学工程研究》[ISSN:1006-6977/CN:61-1281/TN]
- 期数:
-
2021年01期
- 页码:
-
28-32
- 栏目:
-
- 出版日期:
-
2021-03-25
文章信息/Info
- Title:
-
Research on pneumonia image recognition based on deep feature fusion
- 文章编号:
-
1672-6278 (2021)01-0028-05
- 作者:
-
冯翔1△; 康文清2; 吴瀚1; 王风云1; 王星皓1; 季超1
-
1.潍坊医学院生命科学与技术学院,潍坊 261000;2.潍坊市益都中心医院,潍坊 262500
- Author(s):
-
FENG Xiang1; KANG Wenqing2; WU Han1; WANG Fengyun1; WANG Xinghao1; JI Chao1
-
1.College of Life Science and Technology,Weifang Medical College,Weifang 261000,China;2.Weifang Yidu Central Hospital, Weifang 262500
-
- 关键词:
-
深度学习; 跨层连接; 深度融合; 特征识别; 医学影像
- Keywords:
-
Deep learning; Cross-layer connection; Deep fusion; Feature recognition; Medical image
- 分类号:
-
R318
- DOI:
-
10.19529/j.cnki.1672-6278.2021.01.05
- 文献标识码:
-
A
- 摘要:
-
在新型冠状病毒肺炎(COVID-19)疫情背景下,肺炎影像快速准确诊断显得尤为重要。针对肺炎影像纹理及细粒度特征受噪声影响大、常规手段识别率低等问题,本研究构建了一种新的基于跨层连接机制的多主干网络特征融合卷积模型。依托并行特征挖掘思路,利用多尺度感受野挖掘融合来捕获医学图像的局部细节,实现对COVID-19医学影像的筛查,提高诊断准确率。实验表明,本研究模型应用于COVID-19的X光数据集及CT数据集的识别率达到95%以上,对准确、高效诊断新型冠状病毒肺炎具有重大意义。
- Abstract:
-
The rapid and accurate diagnosis of pneumonia images is particularly improtant in the context of the COVID-19 epidemic.As the texture and fine-grained features of pneumonia images greatly affected by noises,and the low recognition rate of conventional methods, we proposed a new multi-backbone network feature fusion convolution model based on the cross-layer connection mechanism. Via the parallel feature mining ideas,we used the multi-scale receptive field mining and fusion to capture the local details of medical images,achieved the medical images screening of COVID-19, and improved the accuracy of diagnosis.Simulations show that the recognition rate of this model applied to COVID-19 X-ray dataset and CT dataset is over 95%, which has great significance for the rapid, accurate and efficient diagnosis of COVID-19.
备注/Memo
- 备注/Memo:
-
(收稿日期:2020-06-18)山东省自然科学基金资助项目(ZR2019BF037)。△通信作者Email:fengxiang230316@163.com
更新日期/Last Update:
2021-04-13