Early detection of soybean mosaic virus using portable Raman spectroscopy coupled with machine learning.

Area of Science:

• Plant pathology
• Spectroscopy
• Artificial intelligence

Background:

• Soybean mosaic virus (SMV) significantly impacts global soybean yield and quality.
• Accurate and early detection of SMV is crucial for effective disease management and yield preservation.

Purpose of the Study:

• To develop a non-invasive method for early detection of SMV using portable Raman spectroscopy and AI.
• To compare the efficacy of different AI models in classifying SMV infection stages.

Main Methods:

• Collected Raman spectra from soybean leaves at various infection stages (0-6 days post-inoculation).
• Preprocessed spectral data using Savitzky-Golay smoothing and Air-PLS baseline correction.
• Developed and evaluated four AI models (1D-CNN, SVM, KNN, BP-ANN) for classification.

Main Results:

• Significant changes in carotenoid levels and distinct spectral responses were observed during SMV infection.
• The 1D-CNN model achieved 90% prediction accuracy.
• Raman spectroscopy advanced SMV detection to 4 days post-inoculation, significantly faster than conventional methods (7-10 days).

Conclusions:

• Raman spectroscopy effectively monitors SMV infection dynamics and reduces diagnosis time.
• The combination of Raman spectroscopy and deep learning is feasible and efficient for in situ early detection of plant viral diseases.
• This approach offers a promising non-destructive method for diagnosing early-stage foliar infections.