摘要
目的:基于互信息粒子群优化(PSO)-长短期记忆(LSTM)神经网络构建医疗设备运行质量预测模型,辅助慢性呼吸系统疾病诊疗设备智能管理.方法:采集设备基本数据、使用数据、维修数据和性能数据进行去噪和标准化处理,构建基于PSO-LSTM神经网络医疗设备运行质量预测模型(简称PSO-LSTM模型),制定设备使用、维护、维修及报废的智能管理方案.选取2019年8月至2023年7月新疆维吾尔自治区人民医院呼吸科临床在用的139台医疗设备,将2019年8月至2021年7月的67台设备采用经验管理模式,2021年8月至2023年7月的72台设备采用智能管理模式.计算传统循环神经网络(RNN)、LSTM神经网络模型训练集和测试集与PSO-LSTM神经网络模型的预测准确性,对比两种管理模式设备管理质量和设备使用操作与技术保障人员以及患者或家属对两种管理模式的管理满意度.结果:PSO-LSTM模型训练集预测准确性的平均绝对百分比误差(MAPE)值和均方根差(RMSE)值分别为0.014和0.008,测试集分别为0.032和0.018,均低于RNN和LSTM模型.采用智能管理模式的设备平均故障频次、平均开机率、管理成本平均增幅、平均维护执行率及平均报废合规率分别为(0.99±0.85)次/年、(95.74±2.16)%、(1.72±1.28)%、(96.49±1.97)%和(97.59±1.49)%,平均故障频次和管理成本平均增幅低于经验管理模式,平均开机率、平均维护执行率和平均报废合规率高于经验管理模式,差异有统计学意义(t=3.297、3.469、2.394、4.187、3.503,P＜0.05);设备使用操作与技术保障人员及患者或家属对采用智能管理模式的设备性能、运行质量、管理方式、管理成本以及诊疗效果满意度评分分别为(94.73±1.85)分、(93.38±3.15)分、(93.48±2.02)分、(94.35±2.34)分和(95.14±2.07)分,均高于经验管理模式,差异有统计学意义(t=4.131、3.827、5.716、3.430、3.173,P＜0.05).结论:基于PSO-LSTM神经网络医疗设备运行质量预测模型能更准确地评估设备运行状况,提高医疗设备临床运行质量,改善临床服务满意度.
Abstract
Objective:To construct a prediction model for the operation quality of medical equipment based on mutual information particle swarm optimization(PSO)-long short-term memory(LSTM)neural network to assist the intelligent management of diagnosis and treatment equipment for chronic respiratory diseases.Methods:The basic data,usage data,maintenance data and performance data of equipment were collected for denoising and standardized processing,and a PSO-LSTM prediction model was constructed,and intelligent management plans for equipment use,maintenance,repair and scrapping were formulated.A total of 139 medical equipment in clinical use in the Respiratory Department of the People's Hospital of Xinjiang Uygur Autonomous Region from August 2019 to July 2023 was selected.67 devices from August 2019 to July 2021 adopted the experience management mode,and 72 devices from August 2021 to July 2023 adopted the intelligent management mode.The prediction accuracy of traditional recurrent neural network(RNN),LSTM neural network model training and test set,and PSO-LSTM neural network model were calculated.The equipment management quality of the two management modes and the satisfaction of equipment operators,technical support personnel,patients and their families with the two management modes were compared.Results:The mean absolute percentage error(MAPE)and root mean square error(RMSE)values of the prediction accuracy of the PSO-LSTM model training set are 0.014 and 0.008,respectively,and the test set is 0.032 and 0.018,respectively,both lower than RNN and RMSE.The failure rate,start-up rate,management cost increase,maintenance implementation rate and scrap compliance rate of the intelligent management mode were(0.99±0.85)times/year,(95.74±2.16)%,(1.72±1.28)%,(96.49±1.97)%and(97.59±1.49%),respectively,and the increase of fault frequency and management cost were lower than those of the experience management mode,while the start-up rate,maintenance implementation rate and scrap compliance rate were than those of the experience management mode,the difference was statistically significant(t=3.297,3.469,2.394,4.187,3.503,P<0.05).The satisfaction scores of equipment operators and technical support personnel and patients and their families on the performance,operating quality,management method,management cost and diagnosis and treatment effect of the equipment of the intelligent management mode were(94.73±1.85),(93.38±3.15),(93.48±2.02),(94.35±2.34)and(95.14±2.07),respectively,which were all higher than those of the experience management,the difference was statistically significant(t=4.131,3.827,5.716,3.430,3.173,P<0.05).Conclusion:PSO-LSTM neural network prediction model can more accurately evaluate the operating status of medical equipment,improve the clinical operation quality of medical equipment and improve clinical service satisfaction.
基金项目
新疆维吾尔自治区自然科学基金资助面上项目(2022D01C129)
维普
万方数据