Single image-based detection of bacterial swarming motility via deep learning

We introduce a deep learning-based classifier capable of rapidly predicting bacterial swarming probability from a single image, which offers a faster, fully automated, objective, and quantitative method for assessing bacterial swarming probability. Trained on Enterobacter sp. SM3, our model achieved a sensitivity of 97.44% and a specificity of 100% in blind testing, accurately distinguishing swarming (positive) from swimming (negative) images in new, unseen SM3 samples. Moreover, this model demonstrated strong external generalization when applied to new bacterial species, including Serratia marcescens DB10 and Citrobacter koseri H6, achieving sensitivity over 97% and specificity exceeding 96% in both cases. The high accuracy and robustness of our deep learning model highlight its potential for rapid, on-site, and objective screening of bacterial swarming motility, thereby enhancing early detection and treatment monitoring of swarming-associated diseases such as inflammatory bowel disease (IBD) and urinary tract infections (UTI).