融合坐标注意力机制的YOLOv3肺结节检测算法

doi:10.16180/j.cnki.issn1007-7820.2024.06.001

电子科技 ›› 2024, Vol. 37 ›› Issue (6): 1-7.doi: 10.16180/j.cnki.issn1007-7820.2024.06.001

• 研究论文 • 下一篇

融合坐标注意力机制的YOLOv3肺结节检测算法

王新宇¹, 赵静文¹, 刘翔¹, 石蕴玉¹, 佘云浪²

1.上海工程技术大学电子电气工程学院,上海 201620
2.同济大学附属上海市肺科医院胸外科,上海 200433

收稿日期:2022-04-06 出版日期:2024-06-15 发布日期:2024-06-20
作者简介:王新宇(1997-),女,硕士研究生。研究方向:医学图像处理。
赵静文(1992-),女,博士,讲师。研究方向:医学图像处理。
刘翔(1972-),男,博士,副教授。研究方向:计算机视觉与人工生命。
基金资助:
上海市自然科学基金(19ZR1421500)

YOLOv3 Lung Nodule Detection Based on Coordinate Attention

WANG Xinyu¹, ZHAO Jingwen¹, LIU Xiang¹, SHI Yunyu¹, SHE Yunlang²

1. School of Electronic and Electric Engineering, Shanghai University of Engineering Science, Shanghai 201620, China
2. Department of Thoracic Surgery, Shanghai Pulmonary Hospital, Tongji University, Shanghai 200433, China

Received:2022-04-06 Online:2024-06-15 Published:2024-06-20
Supported by:
Natural Science Foundation of Shanghai(19ZR1421500)

摘要/Abstract

摘要：

肺结节在CT(Computed Tomography)图像中所占像素较少,增加了检测难度。针对肺结节小目标检测问题,文中提出了融合坐标注意力机制的YOLOv3(You Only Look Once version 3)肺结节检测算法。主干网络采用改进YOLOv3,减少残差块数量并引入扩张卷积模块,并可从目标周围感知上下文信息。在特征利用部分引入坐标注意力机制,捕捉肺结节位置、方向和跨通道信息,精确定位肺结节。改进YOLOv3的损失函数,将边界框建模成高斯分布,利用Wasserstein距离来计算两个分布之间的相似度代替IoU(Intersection over Union)度量,提升模型对目标尺度的敏感性。在LUNA16数据集上的结果显示,肺结节检测的平均精度为89.96%,敏感性为95.37%,与主流目标检测算法相比,精度平均提升了11.33%,敏感性平均提升了9.03%。

关键词: 肺结节, YOLOv3, 扩张卷积, 坐标注意力, 小目标检测, 压缩激发网络, CBAM, NWD

Abstract:

Lung nodules occupy fewer pixels in CT(Computed Tomography), which brings great difficulty to detection. In view of small target detection of lung nodule, this study proposes YOLOv3(You Only Look Once version 3) lung nodule detection algorithm based on coordinate attention. The backbone network adopts the improved YOLOv3, reduces the number of residual blocks and introduces the dilated convolution module to sense context information around the target. In the feature utilization, coordinate attention is introduced to capture the position, direction and cross-channel information, so as to locate lung nodules accurately. The loss function of YOLOv3 is improved, the boundary box is modeled as Gaussian distribution. Wasserstein distance is used to calculate the similarity between the two distributions instead of IoU(Intersection over Union), so as to improve the sensitivity of the target scale. The results on LUNA16 show that the average precision is 89.96% and the sensitivity is 95.37%. Compared with mainstream target detection algorithms, the precision and sensitivity of the proposed method are improved by 11.33% and 9.03%, respectively.

Key words: lung nodules, YOLOv3, dilated convolution, coordinate attention, small target detection, squeeze-and-excitation networks, CBAM, NWD

中图分类号:

TP391

王新宇, 赵静文, 刘翔, 石蕴玉, 佘云浪. 融合坐标注意力机制的YOLOv3肺结节检测算法[J]. 电子科技, 2024, 37(6): 1-7.

WANG Xinyu, ZHAO Jingwen, LIU Xiang, SHI Yunyu, SHE Yunlang. YOLOv3 Lung Nodule Detection Based on Coordinate Attention[J]. Electronic Science and Technology, 2024, 37(6): 1-7.

图/表 8

图1

图2

图3

图4

表1

图5

表2

表3

参考文献 20

[1]	Lyu C, Wu N. Management guidelines for primary lung cancer:Chinese standards in clinical practice[J]. Chinese Journal of Cancer Research, 2019, 31(3):419-426.
[2]	Silva A C, De Paiva A C, Nunes R A, et al. 3D shape analysis to reduce false positives for lung nodule detection systems[J]. Medical and Biological Engineering and Computing, 2017, 55(8):1199-1213.
[3]	程顺达, 程颖, 孙士江. 基于机器学习的肿瘤智能辅助诊断方法[J]. 电子科技, 2022, 35(5): 56-59.
	Cheng Shunda, Cheng Ying, Sun Shijiang. Tumor intelligent auxiliary diagnosis method based on machine learning[J]. Electronic Science and Technology, 2022, 35(5):56-59.
[4]	Khosravan N, Bagci U. S4ND:Single-shot single-scale lung nodule detection[C]. Granada: International Conference on Medical Image Computing and Computer-Assisted Intervention,2018:794-802.
[5]	孙华聪, 彭延军, 郭燕飞, 等. 3D多尺度深度卷积神经网络肺结节检测[J]. 中国图象图形学, 2021, 26(7): 1716-1725.
	Sun Huacong, Peng Yanjun, Guo Yanfei, et al. 3D multi-scale deep convolutional neural networks in pulmonary nodule detection[J]. Journal of Image and Graphics, 2021, 26(7): 1716-1725.
[6]	王乾梁, 石宏理. 基于改进YOLOv3的肺结节检测方法[J]. 中国医学物理学杂志, 2021, 38(9): 1179-1184.
	Wang Qianliang, Shi Hongli. Pulmonary nodule detection based on improved YOLOv3[J]. Chinese Journal of Medical Physics, 2021, 38(9):1179-1184.
[7]	李红光, 于若男, 丁文锐. 基于深度学习的小目标检测研究进展[J]. 航空学报, 2021, 42(7): 107-125.
	Li Hongguang, Yu Ruonan, Ding Wenrui. Research development of small object traching based on deep learning[J]. Acta Aeronautica ET Astronautica Sinica, 2021, 42(7):107-125.
[8]	Liu W, Anguelov D, Erhan D, et al. SSD:Single shot multibox detector[C]. Amsterdam: European Conference on Computer Vision,2016:21-37.
[9]	Redmon J, Divvala S, Girshick R, et al. You only look once: Unified,real-time object detection[C]. Seattle: IEEE Conference on Computer Vision and Pattern Recognition, 2016:779-788.
[10]	Redmon J, Farhadi A. YOLO9000:Better,faster,stronger[C]. Honolulu: IEEE Conference on Computer Vision and Pattern Recognition,2017:6517-6525.
[11]	Redmon J, Farhadi A. YOLOv3:An incremental improvement[C]. Salt Lake City: IEEE Conference on Computer Vision and Pattern Recognition,2018:1-6.
[12]	Bochkovskiy A, Wang C Y, Liao H Y M. YOLOv4: Optimal speed and accuracy of object detection[C]. Seattle: IEEE Conference on Computer Vision and Pattern Recognition,2020:1-17.
[13]	Ren S, He K, Girshick R, et al. Faster R-CNN:Towards real-time object detection with region proposal networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(6):1137-1149.
[14]	He K, Gkioxari G, Dollár P, et al. Mask R-CNN[C]. Venice: IEEE International Conference on Computer Vision,2017:2980-2988.
[15]	赵轩, 周凡, 余汉成. 基于改进特征提取及融合模块的YOLOv3模型[J]. 电子科技, 2022, 35(7):40-45.
	Zhao Xuan, Zhou Fan, Yu Hancheng. Improved YOLOv3 model based on new feature extraction and fusion module[J]. Electronic Science and Technology, 2022, 35(7):40-45.
[16]	Hu J, Shen L, Sun G. Squeeze-and-excitation networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2020, 42(8):2011-2023. doi: 10.1109/TPAMI.2019.2913372 pmid: 31034408
[17]	Woo S, Park J, Lee J Y, et al. CBAM:Convolutional block attention module[C]. Munich: The Fifteenth European Conference on Computer Vision,2018:3-19.
[18]	Hou Q, Zhou D, Feng J. Coordinate attention for efficient mobile network design[C]. Nashville: IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021:13708-13717.
[19]	Misra D. Mish:A self regularized non-monotonic neural activation function[EB/OL].(2019-4-2)[2023-1-11] https://arxiv.53yu.com/vc/arxiv/papers/1908/1908.08681v2.pdf.
[20]	Wang J, Xu C, Yang W, et al. A normalized Gaussian Wasserstein distance for tiny object detection[EB/OL]. (2022-6-14)[2023-1-11] https://arxiv.53yu.com/abs/2110.13389.

网络模型	平均精度/%	敏感性/%
YOLOv3	79.86	85.79
改进基础结构的YOLOv3	82.81	87.56
改进基础结构+Mish激活函数	83.98	88.72
改进基础结构+CA模块	86.97	90.66
改进基础结构+NWD损失函数	86.67	92.45
改进基础结构+Mish+CA	86.84	91.17
改进基础结构+Mish+NWD	87.42	92.64
改进基础结构+CA+NWD	89.17	94.96
本文方法	89.96	95.37

网络模型	平均精度/%	敏感性/%
SE注意力机制	86.25	92.76
CBAM注意力机制	87.97	93.45
CA注意力机制	89.96	95.37

网络模型	平均精度/%	敏感性/%
SSD	74.39	83.74
YOLOv4	78.32	84.65
YOLOv3	79.86	85.79
Faster RCNN	81.97	91.18
本文方法	89.96	95.37

融合坐标注意力机制的YOLOv3肺结节检测算法

YOLOv3 Lung Nodule Detection Based on Coordinate Attention

RichHTML

PDF (PC)

赞

可视化

摘要/Abstract

引用本文

使用本文

图/表 8

参考文献 20

相关文章 4

Metrics

本文评价

推荐阅读 0

[1]	胡永阳,李淼,孟凡开,张峰,孟艺薇,宋宇鲲. 基于Tiny-YOLOv3的网络结构化压缩与加速[J]. 电子科技, 2023, 36(8): 43-48.
[2]	赵轩,周凡,余汉成. 基于改进特征提取及融合模块的YOLOv3模型[J]. 电子科技, 2022, 35(7): 40-45.
[3]	纪玲玉,高永彬,赵呈陆,汤先华,徐凯成,徐嘉诚. 基于3D全卷积网络的腹部动脉CTA分割算法[J]. 电子科技, 2022, 35(3): 38-44.
[4]	叶飞, 刘子龙. 基于改进YOLOv3算法的行人检测研究[J]. 电子科技, 2021, 34(1): 5-9.