基于足底压力的老年人跌倒智能预警模型的构建  

Construction of intelligent fall warning model for the elderly based on plantar pressure

作　　者：刘鑫 张浩哲 何永琴 顾文华[3] 张玉梅 Liu Xin;Zhang Haozhe;He Yongqin;Gu Wenhua;Zhang Yumei(Department of Orthopedics,the Second Affiliated Hospital,PLA Ground Force Medical University,Chongqing 400037,China;School of Automation,Chongqing University,Chongqing 400044,China;School of Electronic Engineering and Optoelectronic Technology,Nanjing University of Science and Technology,Nanjing 210094,China)

机构地区：[1]中国人民解放军陆军军医大学第二附属医院骨科,重庆400037 [2]重庆大学自动化学院,重庆400044 [3]南京理工大学电子工程与光电技术学院,南京210094

出　　处：《中国实用护理杂志》2023年第29期2251-2256,共6页Chinese Journal of Practical Nursing

基　　金：江苏省重点社会发展面上项目(BE2018728);中国人民解放军陆军军医大学临床医学科研人才培养计划(2018XLC3047)。

摘　　要：目的搭建卷积神经网络预测模型,实现对老年人跌倒风险的精准预测。方法2019年6月至2020年2月,采用分层随机整群抽样,利用Python的Matlabplot库和Opencv库对1093例来自南京、重庆共计11所医疗机构、社区及养老机构的老年受试者的足底压力数据进行数据图像化,压缩与剪裁,灰度化、高斯模糊等数据预处理后,将数据划分为训练集983例及验证集110例,用训练集数据训练卷积神经网络模型,用验证集验证模型,并采取ReLU函数抑制过拟合,得到最终的预测模型。结果跌倒预警模型对验证集预测跌倒的敏感度为91.2%,特异度为81.4%,精度为91.5%,AUC为0.865。结论跌倒预测模型对老年人跌倒的预测性能尚可,能对特定场景下的老年人跌倒风险进行预测,在后续改进中应综合改进软硬件构建,从而进一步提升预测准确性。Objective To realize the accurate prediction of the fall risk of the elderly by a convolutional neural network prediction model.Methods Stratified random cluster sampling was used from June 2019 to February 2020,Python′s Matlabplot library and Opencv library were used to perform data preprocessing on the plantar pressure data of 1093 subjects who had come from medical institutionsand community or elderly care institutions of Chongqing and Nanjing,such as data visualization,compression and clipping,grayscale,Gaussian blur,etc.,and then the data were divided into training set(983 cases)and verification set(110 cases),the training set data were used to train the convolutional neural network model,the verification set was used to verify the model,and the ReLU function was used to suppress overfitting to obtain the final prediction model.Results The sensitivity of the fall warning model to the validation set for predicting falls was 91.2%,the specificity was 81.4%,the accuracy was 91.5%,and the AUC was 0.865.Conclusions The fall prediction model can predict the fall risk of the elderly in a specific scenario.In the subsequent improvement,the software and hardware construction should be comprehensively improved to further improve the prediction accuracy.

关 键 词：意外跌倒 一级预防 深度学习 预警模型 

分 类 号：R473[医药卫生—护理学]