Geographical origin and cultivar differentiation of Kava (Piper methysticum) using Artificial Neural Network with FTIR Spectroscopy: A Novel Method

This study presents a novel method for authenticating the geographical origin and cultivar of kava (Piper methysticum) by combining Fourier Transform Infrared (FTIR) spectroscopy with Artificial Neural Networks (ANN). A spectral database of kava varieties from four (4) countries in the Pacific Island region, namely Vanuatu, Fiji, Papua New Guinea, and Hawaii, was used for regional authentication. For samples collected within Vanuatu, spectral data were obtained from the acetone extract of both fresh and dried kava. The ANN predictive model was trained on geographical origin (countries or islands of origin), quality (noble vs tudei), and between different cultivars. ANN achieved near-perfect performance, with a generalized R-Square of 0.99 (training), 0.84 (validation), and 0.95 (test) for geographical origin prediction. Class-specific accuracy was 100% for Vanuatu, Papua New Guinea, and Hawaii. The model achieved near-perfect internal prediction accuracy (R² = 0.99) under repeated-measure validation. Due to the limited availability of independent Fiji samples, no blind-test validation was performed for that region. Thus, robustness claims apply only to regions with sufficient validation data. Significantly, the model demonstrated perfect classification (100% accuracy) for Malo and Santo Island kava samples, highlighting its ability to authenticate micro-regional origins within Vanuatu. For variety differentiation, ANN achieved 100% accuracy for noble versus tudei cultivars, ensuring compliance with Vanuatu’s noble-only export policy. ATR-FTIR spectra of fresh and dried kava acetone extracts exhibited visually distinct patterns among kava cultivars at spectral regions of 1750 cm-1 to 1525 cm-1 and 1124 cm-1 to 900 cm-1, indicating potential for direct differentiation. Visual detection of kava adulteration at 1585 cm-1 was feasible at 1% tudei or wichmannii substitution without advanced analysis. These findings position ANN-FTIR as a rapid, non-destructive, and cost-effective solution for food authentication, geographical indication labeling, and export certification, supporting international standards such as Codex Alimentarius and International Standards Organization (ISO) guidelines.