Phenolic content discrimination in Thai holy basil using hyperspectral data analysis and machine learning techniques.

Area of Science:

• Agricultural science
• Plant biochemistry
• Spectroscopy

Background:

• Traditional antioxidant analysis of holy basil is time-consuming.
• Hyperspectral imaging offers a rapid, non-destructive method for plant property assessment.
• Phytochemical quantification, including phenolic content, is crucial for understanding plant properties.

Purpose of the Study:

• To develop a rapid, non-destructive method for determining total phenolic content in Thai holy basil using hyperspectral imaging and machine learning.
• To classify phenolic content levels into 'low' and 'normal-to-high' categories.
• To evaluate the performance of a neural network model against other machine learning techniques.

Main Methods:

• Acquired hyperspectral data from 26 holy basil cultivars across different growth stages.
• Extracted 22 statistical features in time and frequency domains from spectral data.
• Developed and validated a neural network model for phenolic content classification.

Main Results:

• The neural network model achieved an area under the receiver operating characteristic curve of 0.8113.
• The model demonstrated superior performance compared to other machine learning techniques.
• The model showed increased confidence in predicting phenolic content for older holy basil samples.

Conclusions:

• Hyperspectral imaging integrated with feature extraction and machine learning provides an effective tool for rapid, non-destructive phenolic content assessment in holy basil.
• This approach has the potential to streamline the screening of antioxidant properties in plants.
• The study facilitates efficient decision-making for researchers through comprehensive spectral data analysis.