摘要
目的为解决医疗影像设备故障特征隐蔽性高,单一支持向量机(Support Vector Machine,SVM)方法难以有效提取深层次故障特征,导致故障分类Kappa系数较低的问题,提出基于粒子群优化SVM的CT医疗影像设备故障分类方法。方法采用多类型传感器实时监测CT医疗影像设备的运行状态信号;通过小波变换对采集到的信号进行处理,以去除噪声并提取有效特征,再将处理后的信号输入到深度置信网络中;深度置信网络通过逐层叠加受限玻尔兹曼机,并经历无监督训练和有监督参数调整2个阶段,以精准捕捉并学习CT医疗影像设备故障的深层次特征;将提取到的故障特征输入到基于粒子群优化SVM的多分类器中,通过训练好的模型实现CT医疗影像设备故障分类。结果实验结果表明,该方法在深度置信网络层数为4,输入、隐藏、输出层神经元数量分别为80、150、80的情况下,在G2数据集中的F1得分为0.925,Kappa系数为0.895,海明距离低于0.053。结论本文所提方法可实现CT医疗影像设备故障精准分类,为医疗影像设备的故障诊断和维护提供有力的技术支持。
Objective To solve the problem that the fault features of medical imaging equipment are highly concealed,and the single support vector machine(SVM)method is difficult to effectively extract deep-level fault features,resulting in a low Kappa coefficient for fault classification,to propose a fault classification method for CT medical imaging equipment based on particle swarm optimization SVM.Methods Multi-type sensors were used to monitor the running state signals of CT medical imaging equipment in real time.The collected signals were processed through wavelet transform to remove noise and extract effective features,and then the processed signals were input into the deep belief network.By stacking restricted Boltzmann machine layer by layer and going through two stages of unsupervised training and supervised parameter adjustment,the deep characteristics of CT medical imaging equipment faults were captured and learned by deep belief network accurately.The extracted fault features were input into the multi-classifier based on particle swarm optimization SVM,and the fault classification of CT medical imaging equipment was achieved through the trained model.Results The experimental results showed that the F1 score was 0.925,the Kappa coefficient was 0.895,and the Hamming distance was lower than 0.053 on the G2 dataset when the number of deep confidence network layers was 4 and the number of neurons in input,hidden and output layers was 80,150 and 80 respectively.Conclusion The method proposed in this paper can achieve precise classification of faults in CT medical imaging equipment,providing strong technical support for the fault diagnosis and maintenance of medical imaging equipment.
作者
袁江
刘安琪
杨珍珍
杭国龙
胡云鹏
YUAN Jiang;LIU Anqi;YANG Zhenzhen;HANG Guolong;HU Yunpeng(Department of Medical Engineering,Wuxi Hospital of Traditional Chinese Medicine,Wuxi Jiangsu 214000,China;Department of Medical Imaging,Wuxi Hospital of Traditional Chinese Medicine,Wuxi Jiangsu 214000,China)
出处
《中国医疗设备》
2025年第9期33-38,64,共7页
China Medical Devices
基金
江苏省优势学科建设工程项目(YSHL2103-334)
无锡市科技发展资金项目(Y20212009)。
关键词
粒子群优化算法
支持向量机
CT医疗影像设备
传感器
深度置信网络
惯性权重
particle swarm optimization algorithm
support vector machine(SVM)
CT medical imaging equipment
sensor
deep belief network
inertia weight
作者简介
袁江,邮箱:a719019034@163.com。
