Early infection detection of apple valsa canker pathogens based on SERS

Apple valsa canker has caused the bark rot, and even seriously suffered the growth of fruit trees. Timely and accurately detection of pathogenic infestation is of great significance for the early disease warning, diagnosis, prevention and control measures. Surface-enhanced Raman spectroscopy (SERS) is very feasible in the detection of plant diseases, due mainly to its simple operation, high detection efficiency, and signal strength far beyond the conventional Raman technology. In this research, an accurate and rapid detection was performed on the early infection of apple valsa canker pathogens using SERS technology. The research objects were taken as the rotten bacterial filaments, the healthy and infected apple branches. S-G smoothing and adaptive iterative re-weighted penalized least squares (Air-PLS) were combined to carry out the de-noising and baseline correction of Raman spectra. The characteristic bands of apple branches and pathogenic were identified. Specifically, the healthy apple branches shared the outstanding characteristic peaks near 1 286 and 1 587 cm^-1, whereas, the characteristic peaks of infected apple branches were near 1 365, 1 595, and 2 925 cm^-1. The hypha of pathogenic bacteria had the outstanding characteristic peaks near 731, 1 327, 1 598, and 2 930 cm^-1. At the same time, 15 groups of healthy, infected, and hyphal samples were selected, and found that the peak values of the characteristic peaks near 1 286 and 1 587 cm^-1 of healthy apple branches were less than 0.5 on average. Meanwhile, the peak ratio of the characteristic peaks near 1 365 and 1 595 cm^-1 of infected apple branches was greater than 0.5 in most cases, while the peak ratio of the characteristic peaks near 1 327 and 1 598 cm^-1 of hypha was greater than 1 on average. In the vicinity of 1 280-1 380 cm^-1, the difference between 1 286 and 1 365 cm^-1 was 79, corresponding to the Raman peak generated by healthy and infected samples, while the difference between 1 327 and 1 365 cm^-1 was 38, corresponding to the Raman peak generated by hypha samples and infected samples. The displacement ratio was close to 2:1. The peak intensity of infected apple branches near 1 595 cm^-1 was greater than that of healthy apple branches near 1 587 cm^-1 and that of hypha near 1 598 cm^-1. In general, there was the significant difference in the Raman spectra of healthy apple branches, infected apple branches, and hyphal samples. Back propagation artificial neural network (BP-ANN) was used to construct a qualitative discriminant model, while the BP binary classification model was constructed for the healthy and infected samples, where the recognition rate was up to 96.0%. The BP-four-classification model was also constructed for the healthy and samples with three degrees of infection, where the recognition rate reached 92.0%. In addition, the baseline prediction curve was contained the Raman spectral features in the process of baseline correction. BP-ANN binary and quadruple classification models were then constructed for the baseline prediction curve, where the accuracy was above 90%. Therefore, the SERS combined with BP-ANN can be expected to rapidly and accurately carry out the early diagnosis of apple valsa canker pathogen hypha infection. The finding can provide a technical idea for the early and rapid detection of plant diseases.