论文部分内容阅读
目的探讨使用近红外光谱结合支持向量机法区分耐甲氧西林金葡菌(MR SA)和甲氧西林敏感金葡菌(MSSA)的可行性。方法制作MRSA和MSSA的浓度标准曲线。扩增待测细菌,并根据公式制备相同浓度菌液。采集菌液样本近红外光谱数据,并对数据进行一阶求导、平滑去噪、归一化和基线校正等预处理。根据两种细菌光谱曲线的相关性,对900~2200nm波段数据进行主成分分析。依据累计贡献率结果,选择前三个主成分作为支持向量机的输入向量,分别使用线性、多项式、径向基3种核函数进行建模,比较不同模型区分MSSA和MRSA的准确性。结果 MRSA和MSSA预处理后的光谱曲线相关系数为1.000,两者高度相似。使用主成分处理并采用3种支持向量机核函数建模后,模型的训练和测试准确率均高于95%,其中采用径向基核函数分类结果最好,训练准确率为99.72%±0.21%,测试准确率为99.47%±0.00%。结论使用近红外光谱结合支持向量机的分析方法具有精确区分MRSA和MSSA的能力。
Objective To explore the feasibility of using near-infrared spectroscopy and support vector machine to distinguish MRSA and methicillin-resistant Staphylococcus aureus (MSSA). Methods The concentration standard curve of MRSA and MSSA was made. Amplify the bacteria to be tested and prepare the same concentration of bacteria solution according to the formula. Near-infrared spectroscopy data of the bacterial samples were collected and the data were first-order derivative, smoothing, noise reduction, normalization and baseline correction. Based on the correlation between the two bacterial spectral curves, the principal component analysis was performed on the data from 900 to 2200 nm. According to the results of cumulative contribution rate, the first three principal components are selected as input vectors of support vector machine, and the three kernel functions of linear, polynomial and radial basis are used respectively to model the accuracy of MSSA and MRSA. Results The correlation coefficient of spectral curve after MRSA and MSSA pretreatment was 1.000, which were highly similar. After the principal component processing and the kernel function modeling of three support vector machines were used, the training and testing accuracy of the model were both higher than 95%. The radial basis function was the best for classification and the training accuracy was 99.72% ± 0.21 %, The test accuracy rate is 99.47% ± 0.00%. Conclusion The analytical approach using NIRS with support vector machines has the ability to accurately distinguish between MRSA and MSSA.