Estimating Total Polyphenol Content of Ilex paraguariensis Water Extracts Using Colorimetric Methods and Machine Learning

Ilex paraguariensis contains numerous bioactive compounds and is a form of social interaction in many countries. Yerba mate beverages generally contain higher total polyphenol content (TPC) compared to other plant-based drinks, with hot water-extracted mate offering a higher amount of polyphenol intake, highlighting its importance as a source of antioxidants. For TPC quantification, the colorimetric method using smartphones combined with the Folin-Ciocalteu (FC) assay has been employed for a low-cost and rapid evaluation in different products. However, this approach relies on the FC assay, limiting this method’s accessibility due to the reagent costs and equipment involved. Following this rationale, the objective of this study was to use machine learning models to correlate the color of hot water extracts of Ilex paraguariensis with total polyphenol content. The color must be obtained using a smartphone before starting the FC assay. Thirty samples of Ilex paraguariensis were subjected to three successive hot water extractions. Using a smartphone camera and a color box, the RGB and HSV color channels of the extracts were collected and used in various machine-learning models. While the Elastic Net achieved the best performance (a Relative Root Mean Square Error of 128.53 mg Gallic Acid Equivalent/L), the models' accuracy was constrained by the limited dataset. This approach offers a cost-effective and accessible alternative for TPC estimation, with potential applications in beverage quality control.