Research on data-driven rapid nondestructive quality evaluation method, Calculus Bovis as an example.

Area of Science:

• Analytical Chemistry
• Chemometrics
• Spectroscopy

Background:

• Traditional methods for analyzing animal-derived products suffer from low accuracy due to spatial heterogeneity and limited portable tool resolution.
• Analyzing complex natural products like Calculus bovis requires precise and on-site methods.

Purpose of the Study:

• To develop a data-driven, rapid, non-destructive framework for analyzing animal-derived products.
• To improve the accuracy and on-site applicability of qualitative and quantitative analysis.
• To address the challenge of spatial heterogeneity in complex natural products.

Main Methods:

• Integration of portable near-infrared (NIR) spectroscopy with multi-location spectral fusion and machine learning.
• Application of low-level (raw spectral stitching) and mid-level (feature stitching) fusion techniques.
• Incorporation of variable selection methods for optimized model performance.

Main Results:

• The optimized linear model achieved a qualitative accuracy of 96.70%.
• The mid-level fusion model demonstrated superior quantitative performance with an R² value of 0.9450 and an RPD value of 2.89.
• The developed framework effectively addressed Calculus bovis heterogeneity.

Conclusions:

• The data-driven framework provides a transferable paradigm for the precise on-site analysis of complex natural products.
• This approach meets the demands of analytical chemistry for both efficacy and precision.
• The study highlights the potential of integrated spectroscopy and machine learning for non-destructive product analysis.