Machine Learning-Assisted\nLiquid Crystal Optical Sensor\nArray Using Cysteine-Functionalized Silver Nanotriangles for Pathogen\nDetection in Food and Water

The challenge of rapid identification of bacteria in\nfood and water\nstill persists as a major health problem. To tackle this matter, we\nhave developed a single-probe liquid crystal (LC)-based optical sensing\nplatform for the differentiation of five common bacterial strains,\nincluding Bacillus cereus, Escherichia coli, Pseudomonas aeruginosa, Staphylococcus aureus, and S. typhimurium, using cysteine-functionalized silver\nnanotriangles as signal enhancers. Unique optical patterns were generated\nfrom the interaction of the samples with the LC interface and captured\nby using a camera under polarized light. Pattern recognition was carried\nout based on image analysis and machine learning (ML) calculations.\nAmong the various ML algorithms trained, Support Vector Machines had\nthe best performance and were able to successfully discern the bacteria\nwith 98.89% accuracy. A linear range of 10–10<sup>6</sup> CFU\nmL<sup>–1</sup> and detection limits of under 10 CFU mL<sup>–1</sup> were attained for all of the strains. The proposed\nmethod was tested with water, juice, and milk samples, and prediction\naccuracies of 95.83, 97.92, and 89.58%, respectively, were obtained.\nThe proposed method offers a simple, cost-efficient solution for bacteria\nrecognition.