支持向量机在监护仪无创血压质量控制中的应用研究

摘要

Objective To use the support vector machine (SVM) to establish a quality control (QC) model for QC evaluation of noninvasive blood pressure (NIBP). Methods Eighty Mindray monitors were randomly selected from our hospital. And the blood pressure values under five groups of given parameters were measured by using Fluke simulator, and 80 x 5 matrix vector was set as input. If the maximum error between measured values and given values was within the range of 10mmHg, the monitors were thought as qualified and marked as 1; otherwise, 0 indicated the monitor was unqualified. The 80 x 1 matrix vector was set as output. 60 sets of data were selected as a training set and a multi-input single-output classification model was established by using SVM. The remaining twenty samples was set as a test set to verify whether the method was feasible according to the classification accuracy. Results The classification accuracy rate of the model was 93.3%. Conclusion The method was feasible with novel, fast and convenient advantages. SVM could be used in the quality control of NIBP.%目的：利用支持向量机建立监护仪无创血压质控模型，用于对监护仪进行质量评估。方法随机抽取我院80台迈瑞监护仪，通过福禄克生命体征模拟仪测得80台监护仪在5组给定参数下的血压值，即80×5的矩阵向量作为输入，若5组测量值与给定值的最大误差在±10 mmHg范围内，则认为监护仪合格，用“1”表示；反之，用“0”表示监护仪不合格，即80×1的矩阵向量作为输出。选择60组数据作为训练集，利用支持向量机建立一个多输入单输出的分类模型；剩余的20个样本作为测试集，通过分类准确率来验证该方法是否具有可行性。结果模型评估准确率达到93.3%。结论该方法具有可行性，即支持向量机可应用于监护仪无创血压的质量控制；该方法具有新颖、快捷方便的优点。