基于目标检测的医院感染规范监测方法研究

Research on the monitoring method of hospital infection standard based on target detection

目的:提出基于目标检测的医院感染规范监测方法,提高医护工作者及管理人员的工作效率。方法:采用“云端-边缘”协同的架构设计,利用基于深度学习的目标检测方法对实时摄像机视频数据中的医护人员穿脱防护服规范以及手卫生执行情况进行检测。结果:采用“云端-边缘”协同的方式进行架构设计,可以更好地进行训练和推理;使用带标注的医院感染规范数据训练集对目标检测模型进行训练,并在测试集进行验证,实验结果显示该方法具有较高的准确率,mAP达94.9%。在真实场景中,准确率为94.4%,基本达到了人肉眼识别的水平。结论:结合深度学习的医院感染规范监测方法可以实时检测感控措施的执行情况,具有重要实际应用价值;利用图像数据增强技术对带标注的医院感染规范数据集进行增强,可解决数据不平衡问题,同时增强模型的鲁棒性。

Objective To propose a monitoring method of hospital infection standard based on target detection,so as to improve the efficiency of medical staff and administrators.Methods The"cloud edge"collaborative architecture design was adopted,and the target detection method based on deep learning was used to detect the wearing and taking off of protective clothing of medical staff and their implementation of hand hygiene in real-time camera video data.Results The"cloud edge"collaborative architecture design allows for better training and reasoning.The target detection model is trained by using the labeled hospital infection standard data training set,and verified in the test set.The experimental results show that this method has a high accuracy,and the mAP is 94.9%.In the real scene,the accuracy is 94.4%,basically reaching the level of human eye recognition.Conclusion The hospital infection standard monitoring method combined with deep learning can detect the implementation of infection control measures in real time,which has important practical value.Using image data enhancement technology to enhance the labeled hospital infection standard dataset can solve the problem of data imbalance and enhance the robustness of the model.