基于运营日志数据的CT扫描工作流程优化与需求预测研究OA

Objective Computed tomography(CT)is a cornerstone of modern medical diagnostics;however,the conflict between high equipment costs and complex operational workflows often leads to inefficiencies and prolonged patient waiting times.This study aims to construct a quantitative evaluation system for workflow efficiency based on operational log data obtained from the internet of medical things(IoMT).By analyzing key factors affecting examination efficiency and accurately predicting future demand,this study seeks to provide a scientific basis for decision-making regarding refined management,differentiated scheduling,and the optimization of medical resources in radiology departments.Methods A retrospective analysis was conducted on operational log data from five CT scanners over a three-month period.First,utilizing a data-driven methodology,a system of key performance indicators(KPIs)was extracted and cleaned,including equipment utilization rate,scanning productivity,mean examination turnaround time(TAT),and mean inter-patient interval time.Subsequently,a comparative analysis of KPIs stratified by examination type was performed to identify primary factors influencing equipment productivity.Finally,using daily scan volume data from May and June as a training set,seasonal autoregressive integrated moving average(SARIMA)and random forest regression models were developed to forecast daily scan volumes for July.Model performance was comprehensively evaluated using root mean square error(RMSE),mean absolute error(MAE),coefficient of determination(R2),and directional accuracy.Results The study included over 74 159 scan records from the five devices.KPI analysis revealed a mean examination time of 1.47 min±0.48 min,a mean examination TAT of 7.76 min±3.6 min,and an average equipment utilization rate of 85.76%±16%.Stratified analysis quantified the significant impact of examination protocol complexity on efficiency;data indicated that the mean TAT for multi-site combined scans was significantly higher than that for single-site scans,and the TAT for contrast-enhanced scans was approximately 1.5 times that of non-contrast scans.Regarding demand forecasting,the random forest model outperformed the SARIMA model by approximately 15%across all evaluation metrics,demonstrating that machine learning models more effectively capture the non-linear characteristics of hospital workflows.Conclusions The KPI system constructed based on CT operational log data effectively quantifies the workflow and identifies complex examinations as the key factor constraining CT operational efficiency.Compared to traditional time-series models,this study demonstrates the high accuracy and potential application of the random forest model in forecasting short-term scanning demand.The findings suggest that healthcare institutions can significantly enhance the overall operational efficiency and service quality of radiology departments by implementing differentiated appointment scheduling strategies and applying machine learning predictive models to dynamically allocate technicians and consumables,thereby reducing patient waiting times.