Smartphone image analysis-based fluorescence detection of tetracycline using machine learning

Abstract

Tetracycline (TC) is vastly used as a veterinary drug, making its detection highly important. We have aimed to develop a rapid detection method for TC. For this, BSA-protected Au/Ag bimetallic nanoclusters (BSA-BMNCs) were synthesized for the detection of TC in water and milk. The interaction of TC with BSA shifted the emission of the BMNCs from red to yellow as concentrations of TC increased. Images of the sensing platform were captured with various smartphones and the color and texture information was extracted to generate training datasets for water and milk samples. The datasets were used to train machine learning (ML) algorithms. A model using bagging and artificial neural networks (R² = 0.994, NRMSE = 0.078) for water samples and one using bagging and decision trees (R² = 0.999, NRMSE = 0.027) for milk samples were developed. This study shows the ability of ML algorithms for the development of rapid sensors for the detection of food analytes.

Keywords: Bimetallic nanoclusters; Fluorescence; Machine Learning; Supervised Learning; Tetracycline.