Efficient Detection Classifiers for Genetically-Modified Golden Rice Via Machine Learning

Abstract

Rice is a staple food for over half of the global population, especially in the Philippines. However, traditional rice lacks essential micronutrients like vitamin A, contributing to widespread Vitamin A Deficiency (VAD). Golden Rice was developed to combat VAD, and this is biofortified with beta-carotene, a precursor of Vitamin A. However, concerns about cross-contamination, food safety, and ethics have emerged. Current GMO detection methods, such as PCR and ELISA, are not ideal for large-scale or on-site use since these are intended to be performed inside laboratory and requires technical expertise.  This study presents a novel machine learning (ML)-based approach for the detection of genetically modified Golden Rice using RGB image data and several classification models as an efficient, rapid, non-destructive method to detect GMO Golden Rice. Two datasets of rice images (340 samples of GMO Golden Rice and 340 samples of Traditional Rice) were processed and split for training and testing (80-20 ratio). This study found that WEKA's Random Tree and MATLAB's Trilayered Neural Network achieved 100% accuracy in detecting GMO Golden Rice, with the fastest computational efficiency in their respective platforms. Additional metrics, such as Precision and Recall, further verified the robustness of these classifiers.  This research lays the foundation for developing portable, field-deployable detection tools to empower farmers and regulators while enhancing consumer trust in GMO labeling. Furthermore, the application of ML to GMO rice detection opens new possibilities for biofortified crop monitoring. Future work may explore integrating additional rice features and GMO varieties, validating the results, and expanding this methodology to other GMO rice variants and hybrid varieties to further enhance detection accuracy and scalability.