Rapid artificial-intelligence-based detection of antimicrobial cationic ion activities against Escherichia coli

초록

We developed an AI-based method for early detection of antimicrobial effects on E. coli. Using You Only Look Once (YOLO)-based instance segmentation, the model identified E. coli microcolonies in microscopy images. Collected images were split into training and validation datasets to train an AI model to detect E. coli microcolonies. After training, the AI model achieved a precision, recall, mAP@0.5, and mAP@0.5:0.95, 0.97, 0.97, 0.99, and 0.78, respectively. The results of a time-dependent colony growth test showed that 2 h was the shortest and optimal preincubation time for the early detection. A standard curve was generated to analyse the correlation between the AI predictions and experimental results. The R2 value of the correlation was 0.9958, indicating that the AI model accurately quantified the E. coli microcolonies. Subsequently, a test dataset of 12 samples was used to compare the AI-predicted and actual values, implying that AI-predicted and actual values differed by less than one on a logarithmic scale. Therefore, the AI model was used to determine the antimicrobial activities of the various cationic ions. The AI model accurately estimated the minimal inhibitory and minimal lethal concentrations for E. coli, indicating the potential of our novel methodology to be applied in microbiology.

키워드