Non-destructive analysis of Ganoderma lucidum composition using hyperspectral imaging and machine learning

Area of Science:

• Analytical Chemistry
• Computational Biology
• Mycology

Background:

• Ganoderma lucidum is a valuable medicinal mushroom.
• Traditional quality assessment methods are complex and costly.
• There is a need for rapid, non-destructive quality testing.

Purpose of the Study:

• To develop a fast, non-invasive method for assessing Ganoderma lucidum quality.
• To predict polysaccharide and ergosterol content using hyperspectral imaging and machine learning.

Main Methods:

• Collected hyperspectral images across visible near-infrared and short-wave infrared ranges.
• Measured ergosterol via HPLC and polysaccharides via the phenol-sulfuric acid method.
• Tested feedforward neural network, extreme learning machine, and decision tree models.

Main Results:

• The extreme learning machine model, optimized with a genetic algorithm and voting, showed excellent predictive performance.
• Achieved R² values of 0.96 for polysaccharides and 0.97 for ergosterol.
• Demonstrated the model's ability to accurately quantify key compounds.

Conclusions:

• Hyperspectral imaging combined with machine learning provides a novel approach for Ganoderma lucidum quality assessment.
• This method is non-destructive and efficient.
• Offers a viable alternative to traditional chemical analyses.