Page 87 - 《中国医疗器械杂志》2026年第2期

Chinese Journal of Medical Instrumentation 2026年 第50卷 第2期

临 床 医 学 工 程

文章编号：1671-7104(2026)02-0201-06

基于运营数据的CT球管生命周期阶段预测研究

【作 者】 奉楠馨，谢思源，樊立天，陈柱，刘麒麟
四川大学华西医院 医学工程科，成都市，610041
【摘 要】 研究旨在提出并验证一种利用短期设备运营数据来预测CT球管所处生命周期阶段的方法。采集某大型三甲
医院5台CT设备3个月的详细运营日志及长期的球管更换记录。 通过Kaplan-Meier生存分析方法与对数秩检
验筛选最佳寿命度量指标，结合热反应方程提取15个量化工作负荷的特征。将寿命预测问题转化为四阶段
分类任务，并使用6种机器学习模型进行训练与评估。总扫描量与总曝光秒数比日历天数更能有效反映球管
的耗损过程。随机森林（RandomForest）模型预测性能最佳，五折交叉验证准确率达89.2%，F 1 分数为
0.892。 特征重要性分析证实，累计能量消耗和使用强度的指标与球管的生命周期阶段高度相关，符合物理
退化规律的模式。利用短期运营数据结合机器学习模型，可以高精度地预测CT球管的生命周期阶段。该方
法为医疗机构实施低成本、高效的预测性维护提供了可靠的决策支持工具。
【关 键 词】 机器学习；CT球管；生命周期
【中图分类号】 R318.6
【文献标志码】 A doi: 10.12455/j.issn.1671-7104.250760
Research on CT X-Ray Tube Life Cycle Stage Prediction Based on
Operational Data

【 Authors 】 FENG Nanxin, XIE Siyuan, FAN Litian, CHEN Zhu, LIU Qilin
Department of Medical Engineering, West China Hospital of Sichuan University, Chengdu, 610041
【 Abstract 】 This study aims to propose and validate a method for predicting the lifecycle stage of CT X-ray tubes
using short-term equipment operational data. Detailed operational logs of five CT scanners over a three-
month period, along with long-term tube replacement records, were collected from a large tertiary A-grade
hospital. The Kaplan-Meier survival analysis method and log-rank test were employed to screen for the
optimal lifespan metric. Combined with thermal reaction equations, 15 features quantifying workload were
extracted. The lifespan prediction problem was transformed into a four-stage classification task, which
was trained and evaluated using six machine learning models. Total scan volume and total exposure
seconds reflected the tube degradation process more effectively than calendar days. The Random Forest
(RandomForest) model achieved the best predictive performance, with a 5-fold cross-validation accuracy
of 89.2% and an F1-score of 0.892. Feature importance analysis confirmed that indicators of cumulative
energy consumption and usage intensity were highly correlated with the tube's lifecycle stage, a pattern
consistent with physical degradation laws. Utilizing short-term operational data combined with machine
learning models allows for high-precision prediction of the lifecycle stage of CT X-ray tubes. This method
provides a reliable decision-support tool for healthcare institutions to implement low-cost and efficient
predictive maintenance.
【Key words】 machine learning, CT X-ray tube, lifecycle

0 引言 是现代医学诊断不可或缺的影像技术，然而，

CT设备的高复杂性使其面临运行中断的风险，其
计算机断层扫描（computer tomography，CT） 核心部件X射线球管的突然失效不仅会导致高昂的

更换成本，更会造成设备计划外停机，直接影响患
收稿日期：2025-10-29
基金项目：国家重点研发计划（2023YFC2414600，2023YFC2414602） 者的诊疗流程，延长等待时间，甚至可能延误危重
作者简介：奉楠馨，E-mail: 601583423@qq.com
通信作者：刘麒麟，E-mail: cd610041@163.com 患者的诊断 [1-3] 。

201