koko体育app
一种基于Faster R-CNN的甲状腺结节超声图像目标检测改进算法
doi:
An Improved Object Detection Algorithm for Thyroid Nodule Ultrasound Image Based on Faster R-CNN
-
摘要:
目的 为提高超声图像中甲状腺结节目标检测精度，提出一种基于Faster R-CNN的甲状腺结节目标检测改进算法。 方法 该算法采用结合了可变形卷积（deformable convolution, DC）的ResNeSt50作为主干网络，提高对形状不规则结节的检测效果。并在主干网络后方引入特征金字塔网络（feature pyramid networks, FPN）和感兴趣区域对齐，前者用于减少甲状腺结节漏检误检现象，后者用于提高小尺寸结节的检测精度。此外，在算法训练的反向传播过程中，使用由锐度感知最小化（sharpness-aware minimization, SAM）改进优化器进行参数更新，提高算法的泛化能力。 结果 实验采用来自徐州医科大学附属医院及南京市第一医院6261张甲状腺超声图像，对改进算法的有效性进行对比评估。实验表明，该算法具有一定的优化效果，最终在测试集的AP50高达97.4%，AP@50:5:95较原始模型也提升了10.0%。与原始模型和现有模型相比改进算法有着更高的检测精度，能更精准地检测甲状腺结节，特别在较低的检测框精度要求下有着较高的召回率。 结论 本研究提出的改进方法是有效的甲状腺结节目标检测算法，能精准地检测出甲状腺结节。 -
关键词:
- 甲状腺结节 /
- 超声图像 /
- Faster R-CNN /
- 目标检测 /
- ResNetSt50
Abstract:Objective To propose an improved algorithm for thyroid nodule object detection based on Faster R-CNN so as to improve the detection precision of thyroid nodules in ultrasound images. Methods The algorithm used ResNeSt50 combined with deformable convolution (DC) as the backbone network to improve the detection effect of irregularly shaped nodules. Feature pyramid networks (FPN) and Region of Interest (RoI) Align were introduced in the back of the trunk network. The former was used to reduce missed or mistaken detection of thyroid nodules, and the latter was used to improve the detection precision of small nodules. To improve the generalization ability of the model, parameters were updated during backpropagation with an optimizer improved by Sharpness-Aware Minimization (SAM). Results In this experiment, 6261 thyroid ultrasound images from the Affiliated Hospital of Xuzhou Medical University and the First Hospital of Nanjing were used to compare and evaluate the effectiveness of the improved algorithm. According to the findings, the algorithm showed optimization effect to a certain degree, with the AP50 of the final test set being as high as 97.4% and AP@50:5:95 also showing a 10.0% improvement compared with the original model. Compared with both the original model and the existing models, the improved algorithm had higher detection precision and improved capacity to detect thyroid nodules with better accuracy and precision. In particular, the improved algorithm had a higher recall rate under the requirement of lower detection frame precision. Conclusion The improved method proposed in the study is an effective object detection algorithm for thyroid nodules and can be used to detect thyroid nodules with accuracy and precision. -
Key words:
- Thyroid nodule /
- Ultrasound images /
- koko体育app: Faster R-CNN /
- Object detection /
- ResNetSt50
-
koko体育app
图 1 改进的Faster R-CNN原理图
Figure 1. Schematꦉic diagram of improved Faster R-CNN
DC: deformable convolution; FPN: feature pyramid networks; RPN: region proposal network.
图 2 ResNeSt Block的总体流程
Figure 2. Overall process of ResNeSt Block
FC: fully connected layer; BN: batch normalization; ReLU: rectified linear unit; CB: Conv+BN+ReLU; $ \oplus $: sum of matrix elements; $ \otimes $: multiplication of elements; c: number of channels for output features; $ c' $: number of complete channels for process characterization; K: number of groups; R: number of splits.
图 3 FPN原理图
Figure 3. Schematic diagram of FPN
w: width of input image; Sn: features of stage n.
图 5 基础模型及不同改进模型的PR曲线图
Figure 5. Precision-recall curves of the b🧔asic model and different improved mo✅dels
A, Precision-recall (PR) curve (AP50); B, PR curve (AP75). The legends for graph B are the same as those in graph A.
图 6 不同优化方法的可视化效果图
Figure 6. ෴ Visual rꦦenderings of different optimization methods
A, D, G, J and M, Labeled images; B, RoI Pooling; C, RoI Align; E, FPN not included; F, FPN included; H, VGG16; I, ResNeSt50; K, ordinary convolution; L, deformable convolution; N, SAM not used; O, SAM used. Red boxes indicate the location of nodule.表 1 基础模型及不同改进模型的平均精度对比
Table 1. Comparison of the average precision bet𒐪ween the basic mod🍌el and different improved models
Model AP50/% AP75/% AP@50:5:95/% Base 94.0 66.8 58.7 RoI Align 93.6 73.0 63.1 RoI Align+FPN 97.1 73.9 64.3 RoI Align+FPN+ResNeSt50 97.4 77.3 66.4 RoI Align+FPN+ResNeSt50+DC 97.9 79.0 67.3 RoI Align+FPN+ResNeSt50+DC+SAM 97.4 81.3 68.7 
下载: 导出CSV
var _hmt = _hmt || []; (function() { var hm = document.createElement("script"); hm.src = "https://hm.baidu.com/hm.js?90c4d9819bca8c9bf01e7898dd269864"; var s = document.getElementsByTagName("script")[0]; s.parentNode.insertBefore(hm, s); })(); !function(p){"use strict";!function(t){var s=window,e=document,i=p,c="".concat("https:"===e.location.protocol?"https://":"http://","sdk.51.la/js-sdk-pro.min.js"),n=e.createElement("script"),r=e.getElementsByTagName("script")[0];n.type="text/javascript",n.setAttribute("charset","UTF-8"),n.async=!0,n.src=c,n.id="LA_COLLECT",i.d=n;var o=function(){s.LA.ids.push(i)};s.LA?s.LA.ids&&o():(s.LA=p,s.LA.ids=[],o()),r.parentNode.insertBefore(n,r)}()}({id:"K9y7iMpaU8NS42Fm",ck:"K9y7iMpaU8NS42Fm"}); koko体育-koko体育app koko体育-koko体育网页版koko体育app koko体育-全站app下载(官网) m6米乐app|下载 m6米乐app|主頁欢迎您!!表 2 不同现有模型和改进模型的平均精度对比
Table 2. Comparison of the average precision between different existing models and𝔉 the improved modওel
Model AP50/% AP75/% AP@50:5:95/% YOLOv3 96.2 64.8 58.3 YOLOX 97.0 76.2 66.4 RetinaNet 95.6 69.4 61.4 TOOD 97.1 64.5 64.5 Swin Transformer 97.4 74.7 64.1 Improved model 97.4 81.3 68.7
下载: 导出CSV
-
[1] VACCARELLA S, FRANCESCHI S, BRAY F, et al. Worldwide thyroid-cancer epidemic? The increasing impact of overdiagnosis. New Engl J Med,2016,375(7): 614–617. doi: [2] HAUGEN B R, ALEXANDER E K, BIBLE K C, et al. 2015 American Thyroid Association management guidelines for adult patients with thyroid nodules and differentiated thyroid cancer: the American Thyroid Association guidelines task force on thyroid nodules and differentiated thyroid cancer. Thyroid,2016,26(1): 1–133. doi: [3] 高明. 甲状腺结节和分化型甲状腺癌诊治指南. 中国肿瘤临床,2012,39(17): 1249–1272. doi: [4] ZHENG Z, SU T, WANG Y, et al. A novel ultrasound image diagnostic method for thyroid nodules. Sci Rep,2023,13(1): 1654. doi: [5] LIN X, ZHOU X, TONG T, et al. A super-resolution guided network for improving automated thyroid nodule segmentation. Comput Meth Prog Bio,2022,227: 107186. doi: [6] XU X, WANG C, GUO J, et al. DeepLN: a framework for automatic lung nodule detection using multi-resolution CT screening images. Knowl-Based Syst,2020,189: 105128. doi: [7] KUWANA R, ARIJI Y, FUKUDA M, et al. Performance of deep learning object detection technology in the detection and diagnosis of maxillary sinus lesions on panoramic radiographs. Dentomaxillofac Rad,2021,50(1): 20200171. doi: [8] 王嘉良, 罗健旭, 刘斌, 等. 基于 R-FCN算法的糖尿病眼底病变自动诊断. 计算机工程与应用,2020,56(4): 109–114. doi: [9] 许德刚, 王露, 李凡. 深度学习的典型目标检测算法研究综述. 计算机工程与应用,2021,57(8): 10–25. doi: [10] REDMON J, DIVVALA S, GIRSHICK R, et al. You only look once: unified, real-time object detection//Proceedings of the IEEE conference on computer vision and pattern recognition. Las Vegas: IEEE, 2016: 779−788. doi: . [11] LIU W, ANGUELOV D, ERHAN D, et al. Ssd: Single shot multibox detector//Computer Vision–ECCV 2016: 14th European Conference, Amsterdam: Springer, 2016: 21−37. doi: . [12] XU J, REN H, CAI S, et al. An improved faster R-CNN algorithm for assisted detection of lung nodules. Comput Biol Med,2023,153: 106470. doi: [13] GIRSHICK R, DONAHUE J, DARRELL T, et al. Rich feature hierarchies for accurate object detection and semantic segmentation//Proceedings of the IEEE conference on computer vision and pattern recognition. Columbus: IEEE, 2014: 580−587. doi: . [14] HE K, GKIOXARI G, DOLLAR P, et al. Mask R-CNN//Proceedings of the IEEE international conference on computer vision. Venice: IEEE, 2017: 2961−2969. doi: . [15] WANG L, YANG S, YANG S, et al. Automatic thyroid nodule recognition and diagnosis in ultrasound imaging with the YOLOv2 neural network. World J Surg Oncol,2019,17(1): 12. doi: [16] ZHANG L, ZHUANG Y, HUA Z, et al. Automated location of thyroid nodules in ultrasound images with improved YOLOV3 network. J Xray Sci Technol,2021,29(1): 75–90. doi: [17] 柯威. 基于深度卷积神经网络的甲状腺乳头状癌超声图像识别的研究. 广州: 广东工业大学, 2018. doi: . [18] 秦丽娜. Cascade Mask R-CNN模型的改进及其在甲状腺良恶性结节识别中的应用研究. 南昌: 南昌大学, 2020. [19] REN S, HE K, GIRSHICK R, et al. Faster R-CNN: towards real-time object detection with region proposal networks. IEEE T Pattern Anal,2017,39(6): 1137–1149. doi: [20] SIMONYAN K, ZISSERMAN A. Very deep convolutional networks for large-scale image recognition//Proceedings of the 3rd International Conference on Learning Representations. San Diego: DBIP, 2014. doi: . [21] HE K, ZHANG X, REN S, et al. Deep residual learning for image recognition//Proceedings of the IEEE conference on computer vision and pattern recognition. Las Vegas: IEEE, 2016: 770−778. doi: . [22] ZHU X, HU H, LIN S, et al. Deformable ConvNets V2: more deformable, better results//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Long Beach: IEEE, 2019: 9308−9316. doi: . [23] KWON J, KIM J, PARK H, et al. ASAM: adaptive sharpness-aware minimization for scale-invariant learning of deep neural networks//International Conference on Machine Learning. PMLR, 2021: 5905−5914. doi: . [24] LI X, WANG W, HU X, et al. Selective kernel networks//Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. Long Beach: IEEE, 2019: 510−519. doi: . [25] 张泽苗, 霍欢, 赵逢禹. 深层卷积神经网络的目标检测算法综述. 小型微型计算机系统,2019,40(9): 1825–1831. doi: [26] 郑英豪. Iou-Net模型的改进及其在甲状腺良恶性结节检测与识别中的研究. 南昌: 南昌大学, 2021. [27] 张静漪, 罗燕, 刘加林, 等. 卷积神经网络识别正常及异常甲状腺超声图像的价值. 四川医学,2021,42(3): 305–309. doi: [28] WU X, TAN G, ZHU N, et al. CacheTrack-YOLO: real-time detection and tracking for thyroid nodules and surrounding tissues in ultrasound videos. IEEE J Biomed Health,2021,25(10): 3812–3823. doi: -
下载:
盛开提取 本文遵循知识共享署名—非商业性使用4.0国际许可协议（CC BY-NC 4.0💝），允许第三方对本刊发表的论文自由共享（即在任何媒介以任何形式复制、发行原文）、演绎（即修改、转换或以原文为基础进行创作），必须给出适当的署名，♉提供指向本文许可协议的链接，同时标明是否对原文作了修改；不得将本文用于商业目的。CC BY-NC 4.0许可协议详情请访问 //creativecommons.org/licenses/by-nc/4.0
单击怎么查看大图
计量
- 文章访问数: 16
- HTML全文浏览量: 2
- PDF下载量: 3
- 被引次数: 0
