人工神经网络模型预测老年人骨密度改善运动处方:开发与实施性研究  

The Prediction of Exercise Prescriptions for Improving Bone Mineral Eensity in Elderly People by Using an Artificial Neural Network Model:Development and Usability Study

作　　者：许春艳[1,2] 张蓝天 肖义然 王东敏 王馨塘[2,4] XU Chunyan;ZHANG Lantian;XIAO Yiran;WANG Dongmin;WANG Xintang(School of Sports Science,Beijing Sport University,Beijing 100084,China;Beijing Engineering Research Center of for Sports Nutrition,Beijing 100084,China;Department of Physical Education Teaching and Research,Peking University,Beijing,100084,China;China Institute of Sport and Health,Beijing Sport University,Beijing 100084,China)

机构地区：[1]北京体育大学运动人体科学学院,北京100084 [2]运动营养北京市高等学校工程研究中心,北京100084 [3]北京大学体育教研部,北京100871 [4]北京体育大学中国运动与健康研究院,北京100084

出　　处：《北京体育大学学报》2023年第12期128-138,共11页Journal of Beijing Sport University

基　　金：国家重点研发计划项目“中国人群精准运动处方研制与运动处方库建设”(项目编号:2018YFC2000604);中央高校基本科研业务费专项资金资助(项目编号:2015SYS06)。

摘　　要：目的:探索并建立一个基于人工神经网络(artificial neural network,ANN)的模型,用以预测和生成改善老年人骨密度的运动处方。方法:基于35篇经筛选的文献建立ANN模型,并以北京市某社区30名60~70岁老年女性为研究对象,根据模型给出的运动处方进行实例验证。采用Bland-Altman散点图和命中率(真实值和预测值的差值在其平均值±1倍标准差和±1.96倍标准差范围内数量占总数的百分比)评价模型的预测准确度;采用均方根误差(RMSE)、平均绝对误差(MAE)和决定系数(R2)评价模型的拟合和预测效果;运用配对样本t检验比较研究对象的运动前后数据差异。结果:1)所建ANN模型的预测准确性较高R2=0.95;2)ANN模型生成的运动方案为:进行为期24周的抗阻运动,每周2次,每次45 min。3)运动干预后,受试者骨密度改善的真实值和神经网络模型预测结果无显著性差异(p>0.05);模型在预测全身骨密度时,命中率分别为81.0%和95.2%;在预测脊柱骨密度时,命中率分别为76.2%和90.5%。结论:采用神经网络进行老年人骨密度改善运动处方的模型是可行且有效的,其预测精度较高,可为制定老年人骨密度改善的运动方案提供参考。Objective:This study aims to develop and validate an Artificial Neural Network(ANN)based model for predicting and generating exercise prescriptions to improving bone density in the elderly people.Methods:We established An ANN model utilizing data from 35 selected studies.Thirty female elderly individuals aged 60-70 years from a community in Beijing were chosen as subjects.The exercise prescriptions generated by the model underwent example validation.The model’s predictive accuracy was evaluated using the Bland-Altman scatter plot and hit rate,calculated as the percentage of predictions within mean±1 standard deviation(SD)and±1.96SD.The model’s fit and predictive performance were evaluated using Root Mean Square Error(RMSE),Mean Absolute Error(MAE),and Coefficient of Determination(R2).Paired sample t-tests were used to compare the differences in bone density before and after exercise intervention.Results:1)The ANN model demonstrated high predictive accuracy(R2=0.95);2)The model prescribed a 24-week resistance exercise regimen,twice per week,and 45 minutes each time.3)After the exercise intervention,there was no significant difference between the true value of bone density improvement and the predicted results of the neural network model(p>0.05);The model achieved hit rates of 81.0%and 95.2%in predicting total body bone density,and 76.2%and 90.5%in predicting spinal bone density,respectively.Conclusion:The ANN model for improving exercise prescription in elderly bone density is feasible and effective,offering high predictive accuracy.It can serve as a valuable reference for developing exercise plans aimed at enhancing bone mineral density in elderly people.

关 键 词：老年人 骨密度 人工神经网络 运动处方 运动方案 

分 类 号：R580[医药卫生—内分泌] G804.2[医药卫生—内科学]