应用激光拉曼光谱判别油菜叶片核盘菌早期侵染

病原物核盘菌侵染油菜植株所引发的油菜菌核病严重制约着油菜产业的发展,及早诊断核盘菌的侵染有助于油菜菌核病的早期防治。病原物一般由侵入点向植物寄主四周扩散形成病斑,而与病原物侵染点不同距离的组织区域可代表病害的不同严重程度。该研究采用激光共聚焦显微拉曼光谱仪在800~2 000 cm-1范围内获取健康和染病油菜叶片的拉曼光谱曲线,接着采用小波变换(wavelet transform,WT)进行拉曼光谱的预处理以去除荧光背景的干扰,然后选择主成分因子(principal components,PC-1和PC-2)以及特征参量(1 006,1 156和1 522 cm-1)进行样本间的聚类分析,最后分别基于主成分因子和拉曼特征参量建立最小二乘支持向量机(least squares support vector machine,LS-SVM)进行菌核病侵染油菜叶片不同阶段的判别分析。结果发现采用基于PC-1主成分,1 156和1 522 cm-1处的拉曼强度建立的LS-SVM判别模型可以得到100%的识别率。研究结果表明,通过判别分析油菜叶片菌核病病斑不同区域处的拉曼光谱可以实现核盘菌侵染油菜叶片的早期判别,这为后续探究植物病害的早期监测以及研发油菜叶片菌核病早期诊断拉曼便携仪提供理论参考。

Application of Raman spectroscopy for early detection of rape sclerotinia on rapeseed leaves

Abstract: Raman spectroscopy technique has been widely used in detecting the physiological information of plants. Due to its unique advantages of simple pre-treating, rapid response, high sensitivity and in-situ, nondestructive detecting, it can be performed to acquire biological information. As an important oil crop, oilseed rape (Brassica napus L.) is widely cultivated all around the world. High energy and protein livestock feed are mainly made from its seeds. It is also partly used as potential raw material in synthesizing biodiesel. However, sclerotinia rot of colza, which is generally caused by fungal pathogens sclerotinia sclerotiorum, seriously limited the development of rapeseed industry. Therefore, early detection of the sclerotinia sclerotiorum infection on rape leaves will helpful to discriminate, prevent and cure the sclerotinia rot of colza on rape plants. In general, plant pathogen spread to the tissue around the invasion point, the different distance between the invasion points on the scab represents different severity of the disease on the host plants. In this study, a total of 90 oilseed rape leaves were collected for this experiment. 90 Raman spectral curves of healthy, mid-infected and severe infected oilseed rape leaves were acquired by confocal micro-Raman spectroscopy in the region of 800-2 000 cm-1. Baseline algorithms were employed to process baseline correction. Wavelet transform based on the time-frequency domain can undertake multi-scale decomposition of Raman spectra, therefore, fluorescent background would be removed by reconstructing the signal without low frequency signal. Herein, wavelet transform was used to remove fluorescence background from the original spectral information. Significant differences at 0.05 level among the three kinds of samples at three characteristic peaks (1 006, 1 156 and 1 522 cm-1) were analyzed by one-way analysis of variance (ANOVA) method. Then, principal component analysis (PCA), which can compress the representation of a collection of vectors, was adopted to process cluster analysis. 93% information of the original data was represented by the first two principal components (PC-1 and PC-2). Then, variables of principal components (PC-1 and PC-2) and characteristic peaks (1 006, 1 156 and 1 522 cm-1) were employed to process cluster analysis. Finally, least squares support vector machine (LS-SVM) model, which is highly adaptive, was established based on the data of principal components and the characteristic parameters to discriminate the infected severity. The results revealed that LS-SVM models based on the PC-1, 1 156 and 1 522 cm-1 provided a discriminative accuracy of 100%, LS-SVM discriminative models based on PC-2 and 1 006 cm-1 with accuracy of 66.7% and 70.0%, respectively. The results proved that Raman spectra analysis on the disease spot is meaningful for the early detection of disease. Meanwhile, this research provided a theoretical reference for portable Raman instrument designing in disease infection''s early detection. What is more, different varieties of oilseed rape plants infected by different diseases would be detected if Raman spectroscopy technique is applied. Raman spectra coupled with molecular analysis provide a promising way in detecting specific pathogen of plants at early infection stage in future study.