傅里叶近红外光谱法快速检测5种常见的食源性致病菌

Rapid detection of 5 common foodborne pathogens by Fourier near infrared spectroscopy

中文关键词:  傅里叶近红外光谱法  食源性致病菌  快速检测  主成分分析  偏最小二乘法判别分析

英文关键词:Fourier near-infrared spectroscopy  food-borne pathogens  rapid detection  principal components analysis  partial least-square method

中文摘要:

      目的 建立傅里叶近红外光谱法快速检测5种常见的食源性致病菌。方法 将5种常见致病菌的标准菌株和分离株分别富集培养, 再经冷冻干燥制备成菌粉, 利用傅里叶近红外光谱仪全波长扫描, 得其原始特征光谱, 由仪器自带OPUS软件进行光谱图预处理, 再对用软件进行主成分分析和偏最小二乘法判别分析, 最后建立起快速鉴别菌种的模型。结果 最终显示偏最小二乘法模型优于主成分分析模型。偏最小二乘法模型提取的特征波段为7506.1~6098.1 cm?1, 选择的预处理方式是一阶求导和扣除一条直线, 其决定系数的平方(R2)为93.14%较为接近1, 交叉验证均方根的值为0.361较为接近零, 预测偏差大于3.82, 正确率达到90%, 模型拟合良好。结论 该方法分析速度快, 产出多, 不用试剂, 不污染坏境, 不破坏样品, 适合在线实时检测。

英文摘要:

      Objective To establish a rapid detection method for 5 common food-borne pathogens by Fourier near-infrared spectroscopy. Methods The standard strains and isolates of 5 common pathogens were enriched and cultured respectively, and then their cell powders were collected by freeze-dry. Their original characteristic spectra were obtained by full-wavelength scanning with Fourier near infrared spectrometer. The spectrogram was pretreated by OPUS software, and the processed data were subdivided into principal components. With partial least squares discriminant analysis, a rapid identification model finally establish for these 5 pathogens. Results The results showed that the partial least-square model was superior to the principal components analysis model. The characteristic band was extracted by the partial least-square model from 7506.1 to 6098.1 cm?1. The prefabrication method adopted first derivation and deduction of a straight line. The square of the coefficient of determination (R2) was 93.14%, which was close to 1. The root mean square error cross validation value was close to zero, residual predictive deviation was greater than 3.82, and the accuracy was 90%, which suggested that the model fitted well. Conclusion This method has the advantages of fast analysis speed, high output, no reagent, no pollution to the environment, no damage to samples, and is suitable for on-line and real-time detection.