Protein and lipid content estimation in soybeans using Raman hyperspectral imaging

Area of Science:

• Agricultural Science
• Analytical Chemistry
• Spectroscopy

Background:

• Traditional methods for determining soybean protein and lipid content (Kjeldahl, Soxhlet extraction) are time-consuming, labor-intensive, and destructive.
• Raman hyperspectral imaging (HSI) presents a non-destructive alternative for rapid chemical component analysis.

Purpose of the Study:

• To develop a high-performance, non-destructive model for estimating soybean protein and lipid content using Raman HSI.
• To visualize the spatial distribution of protein and lipids within soybean seeds.

Main Methods:

• Development of a quantitative model using Partial Least Squares Regression (PLSR) with Raman hyperspectral data.
• Calibration dataset comprised 70% of spectral data; validation dataset comprised the remaining 30%.
• Application of the developed model to generate prediction images of component distribution.

Main Results:

• The Raman HSI-PLSR model achieved high accuracy for protein (R²=0.90) and lipid (R²=0.82) content estimation.
• Low prediction errors were observed: Root Mean Squared Error of Prediction (RMSEP) of 1.27 for protein and 0.79 for lipids.
• Successful generation of prediction images illustrating component distribution within individual soybean seeds.

Conclusions:

• Raman HSI combined with PLSR provides an effective, non-destructive method for quantifying soybean protein and lipid content.
• This approach enables rapid analysis without invasive sample preparation and allows for visualization of component distribution.
• The developed model can accurately predict protein and lipid content from single soybean seeds.