基于改进YOLOv5s的肠镜息肉多分类实时检测方法

doi:10.3969/j.issn.1000-1565.2024.04.011

摘要/Abstract

摘要： 为了在肠镜检查过程中对结直肠息肉进行快速鉴别检测,提出一种基于改进YOLOv5s的肠镜息肉多分类实时检测模型.该模型采用ConvNeXt作为主干网络, 融入SimAM注意力机制提升检测性能,同时在颈部网络中使用基于GSConv的slim-neck模块减少网络参数.为了对模型进行训练和测试,构建了包含1 676张息肉图像并由专业医生标注的结直肠息肉数据集.提出的模型在测试集上的平均精度均值(mAP@0.5)为83.0%,相较于改进前提升8.4%,检测速度达到120帧/s. 此外,模型在边缘侧部署检测速度超过25帧/s.结果表明,改进的YOLOv5s满足临床结肠镜检查对实时性与准确性的要求.

关键词: 息肉, 腺瘤, 检测, YOLOv5s, 实时性

Abstract: To facilitate rapid identification and detection of colorectal polyps during colonoscopy procedures, a real-time multi-class detection model for colonoscopic polyps based on modified YOLOv5s is proposed. This model utilizes ConvNeXt as thebackbone network and incorporates the SimAM attention mechanism to improve detection performance. Additionally, a slim-neck module based on GSConv is employed in the neck network to reduce network parameters. For model training and testing, a colorectal polyp dataset containing 1 676 images annotated by professional doctors was constructed. The proposed model achieves a mean Average Precision(mAP@0.5)of 83.0% on the test set, which is an improvement- DOI:10.3969/j.issn.1000-1565.2024.04.011基于改进YOLOv5s的肠镜息肉多分类实时检测方法薛林雁^1,2,3,李轩昂¹,齐晁仪¹,曹杰¹,张颖¹,艾尚璞¹,杨昆^1,2,3(1.河北大学质量技术监督学院,河北保定 071002;2.计量仪器与系统国家地方联合工程研究中心,河北保定 071002;3.河北省新能源汽车动力系统轻量化技术创新中心,河北保定 071002)摘要:为了在肠镜检查过程中对结直肠息肉进行快速鉴别检测,提出一种基于改进YOLOv5s的肠镜息肉多分类实时检测模型.该模型采用ConvNeXt作为主干网络, 融入SimAM注意力机制提升检测性能,同时在颈部网络中使用基于GSConv的slim-neck模块减少网络参数.为了对模型进行训练和测试,构建了包含1 676张息肉图像并由专业医生标注的结直肠息肉数据集.提出的模型在测试集上的平均精度均值(mAP@0.5)为83.0%,相较于改进前提升8.4%,检测速度达到120帧/s. 此外,模型在边缘侧部署检测速度超过25帧/s.结果表明,改进的YOLOv5s满足临床结肠镜检查对实时性与准确性的要求.关键词:息肉;腺瘤;检测;YOLOv5s;实时性中图分类号:TP391.7 文献标志码:A 文章编号:1000-1565(2024)04-0424-09A real-time multi-class detection method for colonoscopy polyps based on improved YOLOv5sXUE Linyan^1,2,3, LI Xuanang¹, QI Chaoyi¹, CAO Jie¹, ZHANG Ying¹, AI Shangpu¹, YANG Kun^1,2,3(1. College of Quality and Technical Supervision, Hebei University, Baoding 071002, China; 2. National & Local Joint Engineering Research Center of Metrology Instrument and System, Baoding 071002, China; 3. New Energy Vehicle Power System Lightweight Technology Innovation Center of Hebei Province, Baoding 071002, China)Abstract: To facilitate rapid identification and detection of colorectal polyps during colonoscopy procedures, a real-time multi-class detection model for colonoscopic polyps based on modified YOLOv5s is proposed. This model utilizes ConvNeXt as thebackbone network and incorporates the SimAM attention mechanism to improve detection performance. Additionally, a slim-neck module based on GSConv is employed in the neck network to reduce network parameters. For model training and testing, a colorectal polyp dataset containing 1 676 images annotated by professional doctors was constructed. The proposed model achieves a mean Average Precision(mAP@0.5)of 83.0% on the test set, which is an improvement- 收稿日期:2024-03-16;修回日期:2024-05-06 基金项目:河北省自然科学基金资助项目(F2023201069);保定市创新能力提升专项项目(2394G027);河北大学研究生创新项目(HBU2024BS021;HBU2024SS011);河北大学科研创新团队项目(IT2023B07);大学生创新创业训练计划创新训练项目(DC2024376;DC2024381) 第一作者:薛林雁(1981—),女,河北大学副教授,主要从事生物医学图像处理方向研究.E-mail: lyxue@hbu.edu.cn 通信作者:杨昆(1976—),男,河北大学教授,博士生导师,主要从事生物医学图像处理方向研究.E-mail: yangkun@hbu.edu.cn第4期薛林雁等:基于改进YOLOv5s的肠镜息肉多分类实时检测方法河北大学学报(自然科学版) 第44卷of 8.4% compared to the model before modification, with a detection speed of 120 frames per second. Moreover,the model exhibits a detection speed exceeding 25 frames per second when deployed on edge devices. The results demonstrate that the improved YOLOv5s meets the clinical requirementsfor real-time and accurate colonoscopy examinations.

Key words: polyps, adenomatous polyps, object detection, YOLOv5s, real-time

中图分类号:

TP391.7

薛林雁,李轩昂,齐晁仪,曹杰,张颖,艾尚璞,杨昆. 基于改进YOLOv5s的肠镜息肉多分类实时检测方法[J]. 河北大学学报(自然科学版), 2024, 44(4): 424-432.

XUE Linyan, LI Xuanang, QI Chaoyi, CAO Jie, ZHANG Ying, AI Shangpu, YANG Kun. A real-time multi-class detection method for colonoscopy polyps based on improved YOLOv5s[J]. Journal of Hebei University(Natural Science Edition), 2024, 44(4): 424-432.

参考文献

[1] 刘宗超,李哲轩,张阳,等.2020全球癌症统计报告解读[J].肿瘤综合治疗电子杂志, 2021, 7(2):1-14. DOI:10.12151/JMCM.2021.02-01.
[2] RODRIGUEZ-DIAZ E, BAFFY G, LO W K,et al. Real-time artificial intelligence-based histological classification of colorectal polyps with augmented visualization - ScienceDirect[J].Gastrointestinal Endoscopy, 2020.DOI:10.1016/j.gie.2020.09.018.
[3] DOUBENI C A, CORLEY D A, QUINN V P, et al. Effectiveness of screening colonoscopy in reducing the risk of death from right and left colon cancer: a large community-based study[J]. Gut, 2018, 67(2): 291-298. DOI: 10.1136/gutjnl-2016-312712.
[4] REX D K, BOLAND C R, DOMINITZ J A, et al. Colorectal cancer screening: recommendations for physicians and patients from the U.S. Multi-Society Task Force on Colorectal Cancer[J]. Gastrointest Endosc, 2017, 86(1): 18-33. DOI: 10.1016/j.gie.2017.04.003.
[5] ZHAO S B, WANG S L, PAN P, et al. Magnitude, risk factors, and factors associated with adenoma miss rate of tandem colonoscopy: a systematic review and meta-analysis[J]. Gastroenterology, 2019, 156(6): 1661-1674. DOI: 10.1053/j.gastro.2019.01.260.
[6] LUI T K L, GUO C G, LEUNG W K. Accuracy of artificial intelligence on histology prediction and detection of colorectal polyps: a systematic review and meta-analysis[J]. Gastrointest Endosc, 2020, 92(1): 11-22. DOI: 10.1016/j.gie.2020.02.033.
[7] BYRNE M F, SHAHIDI N, REX D K. Will computer-aided detection and diagnosis revolutionize colonoscopy?[J]. Gastroenterology, 2017, 153(6): 1460-1464. DOI: 10.1053/j.gastro.2017.10.026.
[8] 蒋西然,蒋韬,孙嘉瑶,等.深度学习人工智能技术在医学影像辅助分析中的应用[J].中国医疗设备, 2021, 36(6): 164-171. DOI: 10.3969/j.issn.1674-1633.2021.06.040.
[9] OZAWA T, ISHIHARA S, FUJISHIRO M, et al. Automated endoscopic detection and classification of colorectal polyps using convolutional neural networks[J]. Therap Adv Gastroenterol, 2020, 13: 1756284820910659. DOI: 10.1177/1756284820910659.
[10] NOGUEIRA RODRÍGUEZ A, DOMÍNGUEZ CARBAJALES R, CAMPOS TATO F, et al. Real-time polyp detection model using convolutional neural networks[J]. Neural Comput Appl, 2021, 34(13): 1-22.
[11] MO X, TAO K, WANG Q, et al. An efficient approach for polyps detection in endoscopic videos based on faster R-CNN[C] //2018 24th International Conference on Pattern Recognition(ICPR). Beijing, China. IEEE, 2018: 3929-3934. DOI: 10.1109/ICPR.2018.8545174.
[12] PACAL I, KARABOGA D. A robust real-time deep learning based automatic polyp detection system[J]. Comput Biol Med, 2021, 134: 104519. DOI: 10.1016/j.compbiomed.2021.104519.
[13] CAI H, FENG X, YIN R, et al. MIST: multiple instance learning network based on Swin Transformer for whole slide image classification of colorectal adenomas[J]. The Journal of Pathology, 2023, 259(2): 125-135. DOI: 10.1002/path.6027.
[14] 张文明,姚振飞,高雅昆,等.一种平衡准确性以及高效性的显著性目标检测深度卷积网络模型[J].电子与信息学报, 2020, 42(5): 1201-1208. DOI: 10.11999/JEIT190229.
[15] JIANG L, LI X.An efficient and accurate object detection algorithm and its application[J].IEEE, 2020.DOI:10.1109/ITOEC49072.2020.9141752.
[16] ADARSH P, RATHI P, KUMAR M.YOLO v3-Tiny: object detection and recognition using one stage improved model[C] //2020 6th International Conference on Advanced Computing and Communication Systems(ICACCS).IEEE, 2020.DOI:10.1109/ICACCS48705.2020.9074315.
[17] BOCHKOVSKIY A, WANG C Y, LIAO H Y M. YOLOv4: optimal speed and accuracy of object detection[EB/OL]. 2020: arXiv: 2004.10934. http://arxiv.org/abs/2004.10934.
[18] ZHAOYANG N,GUOQIANG Z,HUI Y.A review on the attention mechanism of deep learning[J].Neurocomputing, 2021.DOI:10.1016/j.neucom.2021.03.091. 555/3295222.3295349.
[19] LIU Z, MAO H Z, WU C Y, et al. A ConvNet for the 2020s[C] //2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR). New Orleans, LA, USA. IEEE, 2022: 11966-11976. DOI: 10.1109/CVPR52688.2022.01167.
[20] Yang L, Zhang R-Y, Li L, et al. SimAM: A Simple, Parameter-Free Attention Module for Convolutional Neural Networks[C] //International Conference on Machine Learning. PMLR, 2021: 11863-11874.
[21] LI H L, LI J, WEI H B, et al. Slim-neck by GSConv: a better design paradigm of detector architectures for autonomous vehicles[EB/OL]. 2022: arXiv: 2206.02424. http://arxiv.org/abs/2206.02424.
[22] GEVORGYAN Z. SIoU loss: more powerful learning for bounding box regression. 2022: arXiv: 2205.12740. http://arxiv.org/abs/2205.12740.
[23] CHENG H, ANYUAN Y, HONGLIN H. 2021. Using combined Soft-NMS algorithm Method with Faster R-CNN model for Skin Lesion Detection. In Proceedings of the 6th International Conference on Robotics and Artificial Intelligence(ICRAI 20). Association for Computing Machinery, New York, NY, USA, 5–8. https://doi.org/10.1145/3449301.3449303.
[24] BERNAL J, SÁNCHEZ F J, FERNÁNDEZ-ESPARRACH G, et al. WM-DOVA maps for accurate polyp highlighting in colonoscopy: validation vs. saliency maps from physicians[J]. Comput Med Imaging Graph, 2015, 43: 99-111. DOI: 10.1016/j.compmedimag.2015.02.007. (