Machine Learning-Assisted Liquid Crystal Optical Sensor Array Using Cysteine-Functionalized Silver Nanotriangles for Pathogen Detection in Food and Water

The challenge of rapid identification of bacteria in food and water still persists as a major health problem. To tackle this matter, we have developed a single-probe liquid crystal (LC)-based optical sensing platform for the differentiation of five common bacterial strains, including <i>Bacillus cereus</i>, <i>Escherichia coli</i>, <i>Pseudomonas aeruginosa</i>, <i>Staphylococcus aureus</i>, and <i>S. typhimurium</i>, using cysteine-functionalized silver nanotriangles as signal enhancers. Unique optical patterns were generated from the interaction of the samples with the LC interface and captured by using a camera under polarized light. Pattern recognition was carried out based on image analysis and machine learning (ML) calculations. Among the various ML algorithms trained, Support Vector Machines had the best performance and were able to successfully discern the bacteria with 98.89% accuracy. A linear range of 10-10⁶ CFU mL^-1 and detection limits of under 10 CFU mL^-1 were attained for all of the strains. The proposed method was tested with water, juice, and milk samples, and prediction accuracies of 95.83, 97.92, and 89.58%, respectively, were obtained. The proposed method offers a simple, cost-efficient solution for bacteria recognition.