基于改进单点多盒检测器的麻醉复苏目标检测方法

doi:10.11772/j.issn.1001-9081.2022121917

《计算机应用》唯一官方网站 ›› 2023, Vol. 43 ›› Issue (12): 3941-3946.DOI: 10.11772/j.issn.1001-9081.2022121917

基于改进单点多盒检测器的麻醉复苏目标检测方法

罗荣昊, 程志友, 汪传建(), 刘思乾, 汪真天

安徽大学互联网学院，合肥 230039

收稿日期:2023-01-04 修回日期:2023-04-05 接受日期:2023-04-06 发布日期:2023-04-12 出版日期:2023-12-10
通讯作者: 汪传建
作者简介:罗荣昊（1997—），男，安徽滁州人，硕士研究生，主要研究方向：目标检测、微动作识别
程志友（1972—），男，安徽安庆人，教授，博士，主要研究方向：电能质量分析与控制
刘思乾（1997—），男，安徽巢湖人，硕士研究生，主要研究方向：目标检测
汪真天（1999—），男，安徽铜陵人，硕士研究生，主要研究方向：目标检测、光学文字识别。
基金资助:
国家自然科学基金资助项目(82272225)

Anesthesia resuscitation object detection method based on improved single shot multibox detector

Ronghao LUO, Zhiyou CHENG, Chuanjian WANG(), Siqian LIU, Zhentian WANG

School of Internet，Anhui University，Hefei Anhui 230039，China

Received:2023-01-04 Revised:2023-04-05 Accepted:2023-04-06 Online:2023-04-12 Published:2023-12-10
Contact: Chuanjian WANG
About author:LUO Ronghao， born in 1997， M. S. candidate. His research interests include object detection， micro-action recognition.
CHENG Zhiyou， born in 1972， Ph. D.， professor. His research interests include analysis and control of power quality.
LIU Siqian， born in 1997， M. S. candidate. His research interests include object detection.
WANG Zhentian， born in 1999， M. S. candidate. His research interests include object detection， optical character recognition.
Supported by:
National Natural Science Foundation of China(82272225)

摘要/Abstract

摘要：

麻醉复苏目标检测模型常被用于帮助医护人员检测麻醉病人的复苏。病人复苏时面部动作的目标较小且幅度不明显，而现有的单点多盒检测器（SSD）难以准确实时地检测病人的面部微动作特征。针对原有模型检测速度低、容易出现漏检的问题，提出一种基于改进SSD的麻醉复苏目标检测方法。首先，将原始SSD的主干网络VGG（Visual Geometry Group）16更换为轻量级的主干网络MobileNetV2，并把标准卷积替换成深度可分离卷积；同时，通过对病人照片的特征提取采用先升维再降维的计算方式减少计算量，从而提高模型的检测速度；其次，将SSD提取的不同尺度特征层中融入坐标注意力（CA）机制，并通过对通道和位置信息加权的方式提升特征图提取关键信息的能力，优化网络的定位分类表现；最后，闭眼数据集CEW（Closed Eyes in the Wild）、自然标记人脸数据集LFW（Labeled Faces in the Wild）和医院麻醉病患面部数据集HAPF（Hospital Anesthesia Patient Facial）这3个数据集上进行对比实验。实验结果表明，所提模型的平均精度均值（mAP）达到了95.23%，检测照片的速度为每秒24帧，相较于原始SSD模型的mAP提升了1.39个百分点，检测速度提升了140%。因此，所提模型在麻醉复苏检测中具有实时准确检测的效果，能够辅助医护人员进行苏醒判定。

关键词: 麻醉复苏, 面部特征识别, 单点多盒检测器, MobileNetV2, 注意力机制

Abstract:

The target detection model of anesthesia resuscitation is often used to help medical staff to perform resuscitation detection on anesthetized patients. The targets of facial actions during patient resuscitation are small and are not obvious， and the existing Single Shot multibox Detector （SSD） is difficult to accurately detect the facial micro-action features of patients in real time. Aiming at the problem that the original model has low detection speed and is easy to have missed detection， an anesthesia resuscitation object detection method based on improved SSD was proposed. Firstly， the backbone network VGG （Visual Geometry Group）16 of the original SSD was replaced by the lightweight backbone network MobileNetV2， and the standard convolutions were replaced by the depthwise separable convolutions. At the same time， the calculation method of first increasing and then reducing the dimension of the extracted features from patient photos was used to reduce computational cost， thereby improving detection speed of the model. Secondly， the Coordinate Attention （CA） mechanism was integrated into the feature layers with different scales extracted by the SSD， and the ability of the feature map to extract key information was improved by weighting the channel and location information， so that the network positioning and classification performance was optimized. Finally， comparative experiments were carried out on three datasets： CEW（Closed Eyes in the Wild）， LFW（Labeled Faces in the Wild）， and HAPF（Hospital Anesthesia Patient Facial）. Experimental results show that the mean Average Precision （AP） of the proposed model reaches 95.23%， and the detection rate of photos is 24 frames per second， which are 1.39 percentage points higher and 140% higher than those of the original SSD model respectively. Therefore， the improved model has the effect of real-time accurate detection in anesthesia resuscitation detection， and can assist medical staff in resuscitation detection.

Key words: anesthesia resuscitation, facial feature recognition, Single Shot multibox Detector (SSD), MobileNetV2, attention mechanism

中图分类号:

TP391.41

罗荣昊, 程志友, 汪传建, 刘思乾, 汪真天. 基于改进单点多盒检测器的麻醉复苏目标检测方法[J]. 计算机应用, 2023, 43(12): 3941-3946.

Ronghao LUO, Zhiyou CHENG, Chuanjian WANG, Siqian LIU, Zhentian WANG. Anesthesia resuscitation object detection method based on improved single shot multibox detector[J]. Journal of Computer Applications, 2023, 43(12): 3941-3946.

图/表 13

图1 SSD的结构

Fig.1 Structure of SSD

图2 改进的SSD结构

Fig.2 Improved SSD structure

图3 MobileNetV2卷积结构

Fig.3 MobileNetV2 convolution structure

图4 深度可分离卷积

Fig.4 Depthwise separable convolution

图5 CA模块的结构

Fig.5 Structure of CA module

图6 数据集图像标注

Fig. 6 Dataset image annotation

图7 网络训练结果

Fig.7 Network training results

图8 实际预测结果

Fig. 8 Actual forecast results

表1 不同模型的检测精确度对比 (%)

Tab.1 Comparison of detection precision of different models

模型	不同表情检测精确度				mAP
模型	睁眼	闭眼	张嘴	闭嘴	mAP
SSD	92.48	96.09	94.91	91.87	93.84
CA-SSD	94.78	97.53	95.47	92.33	95.03
MobileNetV2-SSD	93.41	96.85	96.30	93.37	94.99
本文模型	93.75	97.05	96.78	93.35	95.23

表2 不同模型的大小和检测速度对比

Tab.2 Comparison of size and detection speed of different models

模型	模型大小/MB	检测速度/（frame·s^-1）
模型	模型大小/MB	显卡	处理器
SSD	92.1	153	10
CA-SSD	93.4	125	9
MobileNetV2-SSD	15.8	81	25
本文模型	17.3	74	24

图9 不同模型的灵敏度对比

Fig.9 Comparison of sensitivity of different models

表 3 不同模型的平均对数漏检率对比

Tab.3 Comparison of log-average miss rate of different models

模型	不同表情平均对数漏检率
模型	睁眼	闭眼	张嘴	闭嘴
SSD	0.18	0.08	0.09	0.12
CA-SSD	0.13	0.07	0.07	0.13
MobileNetV2-SSD	0.17	0.11	0.05	0.09
本文模型	0.13	0.06	0.04	0.05

图10 受试者特征曲线

Fig.10 Receiver operating characteristic curve

参考文献 24

1	DESBOROUGH J P. The stress response to trauma and surgery［J］. British Journal of Anaesthesia，2000，85（1）：109-117. 10.1093/bja/85.1.109
2	DOBSON G P. Addressing the global burden of trauma in major surgery［EB/OL］.［2022-12-20］. doi：10.3389/fsurg.2015.00043 .
3	CUSACK B， BUGGY D J. Anaesthesia， analgesia， and the surgical stress response［J］. BJA Education， 2020，20（9）： 321-328. 10.1016/j.bjae.2020.04.006
4	HIROSE M， OKUTANI H， HASHIMOTO K， et al. Intraoperative assessment of surgical stress response using nociception monitor under general anesthesia and postoperative complications： a narrative review［J］. Journal of Clinical Medicine， 2022，11（20）： No. 6080. 10.3390/jcm11206080
5	PENSON D F. Re： relationship between occurrence of surgical complications and hospital finances［J］. The Journal of Urology， 2013， 190（6）： 2211-2213. 10.1016/j.juro.2013.08.063
6	DOBSON G P. Trauma of major surgery： a global problem that is not going away ［J］. International Journal of Surgery， 2020， 81：47-54. 10.1016/j.ijsu.2020.07.017
7	LUDBROOK G L. The hidden pandemic： the cost of postoperative complications ［J］. Current Anesthesiology Reports， 2021， 12（1）：1-9. 10.1007/s40140-021-00493-y
8	郭清厚，钟娆霞，莫玉林.靶向预控护理在全麻手术患者复苏期躁动管理中的应用［J］.齐鲁护理杂志，2019，25（6）：92-94. 10.3969/j.issn.1006-7256.2019.06.033
	GUO Q H， ZHONG R X， MO Y L. Application of targeted pre-control nursing in restlessness management of patients undergoing general anesthesia surgery during recovery ［J］. Journal of Qilu Nursing， 2019， 25（6）： 92-94. 10.3969/j.issn.1006-7256.2019.06.033
9	SOUKUPOVÁ T， CECH J. Real-time eye blink detection using facial landmarks ［EB/OL］.［2022-12-20］. .
10	NOUSIAS G， E-K PANAGIOTOPOULOU， DELIBASIS K， et al. Video-based eye blink identification and classification［J］. IEEE Journal of Biomedical and Health Informatics， 2022， 26（7）： 3284-3293. 10.1109/jbhi.2022.3153407
11	DE LA CRUZ G， LIRA M， LUACES O， et al. Eye-LRCN： a long-term recurrent convolutional network for eye blink completeness detection［J/OL］. IEEE Transactions on Neural Networks and Learning Systems， 2022［2022-11-29］. . 10.1109/tnnls.2022.3202643
12	CHEN Y， ZHAO D， HE G. Deep learning-based fatigue detection for online learners［C］// Proceedings of the 2022 5th International Conference on Pattern Recognition and Artificial Intelligence. Piscataway： IEEE， 2022 ： 924-927. 10.1109/prai55851.2022.9904096
13	WANG Z， CHAI J， XIA S. Realtime and accurate 3D eye gaze capture with DCNN-based iris and pupil segmentation ［J］. IEEE Transactions on Visualization and Computer Graphics， 2021，27（1）：190-203. 10.1109/tvcg.2019.2938165
14	PRINSEN V， JOUVET P， OMAR S A， et al. Automatic eye localization for hospitalized infants and children using convolutional neural networks ［J］. International Journal of Medical Informatics， 2021， 146： 104344. 10.1016/j.ijmedinf.2020.104344
15	LIU W， ANGUELOV D， ERHAN D，et al. SSD：single shot multibox detector ［C］// Proceedings of the 2016 European Conference on Computer Vision，LNCS 9905. Cham： Springer，2016： 21-37.
16	SIMONYAN K， ZISSERMAN A. Very deep convolutional networks for large-scale image recognition［EB/OL］.［2022-12-20］. .
17	SANDLER M， HOWARD A， ZHU M， et al. MobileNetV2： inverted residuals and linear bottlenecks ［C］// Proceedings of the 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway： IEEE， 2018：4510-4520. 10.1109/cvpr.2018.00474
18	HU J， SHEN L， ALBANIE S， et al. Squeeze-and-excitation networks.［J］. IEEE Transactions on Pattern Analysis and Machine Intelligence， 2020， 42（8）：2011-2023. 10.1109/tpami.2019.2913372
19	WANG Q， WU B， ZHU P， et al. ECA-Net： efficient channel attention for deep convolutional neural networks［C］// Proceedings of the 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway： IEEE， 2020 ：11531-11539. 10.1109/cvpr42600.2020.01155
20	HOU Q， ZHOU D， FENG J. Coordinate attention for efficient mobile network design［C］// Proceedings of the 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway： IEEE，2021 ：13708-13717. 10.1109/cvpr46437.2021.01350
21	ZHAO L， WANG Z， ZHANG G，et al.Eye state recognition based on deep integrated neural network and transfer learning［J］.Multimedia Tools and Applications， 2018， 77（15）：19415-19438. 10.1007/s11042-017-5380-8
22	HUANG G B， MATTAR M， BERG T，et al.Labeled faces in the wild： a database for studying face recognition in unconstrained environments ［EB/OL］.［2022-12-22］.. 10.1117/12.2080393
23	TAN C， SUN F， KONG T， et al. A survey on deep transfer learning ［C］// Proceedings of the 2018 International Conference on Artificial Neural Networks and Machine Learning. Cham： Springer， 2018：270-279. 10.1007/978-3-030-01424-7_27
24	MA N， ZHANG X.， ZHENG H T，et al. ShuffleNet v2： practical guidelines for efficient cnn architecture design ［C］// Proceedings of the 2018 European Conference on Computer Vision. Cham：Springer， 2018 ：122-138. 10.1007/978-3-030-01264-9_8

[1]	杨昊, 张轶. 基于上下文信息和多尺度融合重要性感知的特征金字塔网络算法[J]. 《计算机应用》唯一官方网站, 2023, 43(9): 2727-2734.
[2]	袁国龙, 张玉金, 刘洋. 基于残差反馈和自注意力的图像篡改取证网络[J]. 《计算机应用》唯一官方网站, 2023, 43(9): 2925-2931.
[3]	王宏, 钱清, 王欢, 龙永. 融合大核注意力卷积的轻量化图像篡改定位算法[J]. 《计算机应用》唯一官方网站, 2023, 43(9): 2692-2699.
[4]	张秋余, 温永旺. 用于语音检索的三联体深度哈希方法[J]. 《计算机应用》唯一官方网站, 2023, 43(9): 2910-2918.
[5]	崔雨萌, 王靖亚, 刘晓文, 闫尚义, 陶知众. 融合注意力和裁剪机制的通用文本分类模型[J]. 《计算机应用》唯一官方网站, 2023, 43(8): 2396-2405.
[6]	齐爱玲, 王宣淋. 基于中层细微特征提取与多尺度特征融合细粒度图像识别[J]. 《计算机应用》唯一官方网站, 2023, 43(8): 2556-2563.
[7]	金泽熙, 李磊, 刘继. 基于改进领域分离网络的迁移学习模型[J]. 《计算机应用》唯一官方网站, 2023, 43(8): 2382-2389.
[8]	梁美佳, 刘昕武, 胡晓鹏. 基于改进YOLOv3的列车运行环境图像小目标检测算法[J]. 《计算机应用》唯一官方网站, 2023, 43(8): 2611-2618.
[9]	王静红, 周志霞, 王辉, 李昊康. 双路自编码器的属性网络表示学习[J]. 《计算机应用》唯一官方网站, 2023, 43(8): 2338-2344.
[10]	段升位, 程欣宇, 王浩舟, 王飞. 基于改进的YOLOv5的大坝表面病害检测算法[J]. 《计算机应用》唯一官方网站, 2023, 43(8): 2619-2629.
[11]	刘源, 董永权, 贾瑞, 杨昊霖. 面向个性化课程推荐的分层分期注意力网络模型[J]. 《计算机应用》唯一官方网站, 2023, 43(8): 2358-2363.
[12]	梁敏, 刘佳艺, 李杰. 融合迭代反馈与注意力机制的图像超分辨重建方法[J]. 《计算机应用》唯一官方网站, 2023, 43(7): 2280-2287.
[13]	叶坤佩, 熊熙, 丁哲. 基于领域融合和时间权重的招工推荐模型[J]. 《计算机应用》唯一官方网站, 2023, 43(7): 2133-2139.
[14]	轩勃娜, 李进, 宋亚飞, 马泽煊. 基于改进MobileNetV2的恶意代码分类方法[J]. 《计算机应用》唯一官方网站, 2023, 43(7): 2217-2225.
[15]	拓雨欣, 薛涛. 融合指针网络与关系嵌入的三元组联合抽取模型[J]. 《计算机应用》唯一官方网站, 2023, 43(7): 2116-2124.

基于改进单点多盒检测器的麻醉复苏目标检测方法

Anesthesia resuscitation object detection method based on improved single shot multibox detector

RichHTML

PDF

可视化

摘要/Abstract

引用本文

使用本文

图/表 13

参考文献 24

相关文章 15

编辑推荐

Metrics