Rapid Authentication of Plant-Based Milk Alternatives by Coupling Portable Raman Spectroscopy With Machine Learning.

Area of Science:

• Analytical Chemistry
• Food Science
• Spectroscopy

Background:

• Plant-based milk alternatives (PBMA) are gaining popularity due to increased lactose intolerance and environmental concerns.
• Rapid authentication methods for verifying the biological origin of PBMA are currently limited.

Purpose of the Study:

• To develop a rapid, on-site analytical method for authenticating and identifying PBMA from six different plant species.
• To utilize a portable Raman spectrometer combined with machine learning algorithms for this purpose.

Main Methods:

• Prepared unprocessed (blended raw) and processed (filtered, pasteurized) PBMA in a laboratory setting.
• Collected Raman spectra without sample preparation.
• Tested and compared three machine learning algorithms: k-nearest neighbor (KNN), support vector machine (SVM), and random forest (RF).

Main Results:

• The random forest (RF) algorithm demonstrated the highest performance in identifying plant sources for unprocessed PBMA, achieving over 95% accuracy.
• RF models also achieved high accuracy (over 94%) in identifying species for processed PBMA when combining spectra from both unprocessed and processed samples.
• The portable Raman spectrometer effectively captured chemical fingerprints for plant species identification.

Conclusions:

• The developed Raman spectroscopic method coupled with machine learning provides a rapid (within 5 minutes) and reliable tool for PBMA authentication.
• This non-destructive technique can ensure the authenticity of the biological origin of various PBMA.
• The method offers inspection laboratories a valuable screening tool for verifying PBMA authenticity.