Classification of Microbial Activity and Inhibition Zones Using Neural Network Analysis of Laser Speckle Images

This study addresses the challenge of rapidly and accurately distinguishing zones of microbial activity from antibiotic inhibition zones in Petri dishes. We propose a laser speckle imaging technique enhanced with subpixel correlation analysis to monitor dynamic changes in the inhibition zone surrounding an antibiotic disc. This method provides faster results compared to the standard disk diffusion assay recommended by EUCAST. To enable automated analysis, we used machine learning algorithms for classifying areas of bacterial or fungal activity versus inhibited growth. Classification is performed over short time windows (e.g., 1 h), supporting near-real-time assessment. To further improve accuracy, we introduce a correction method based on the known spatial dynamics of inhibition zone formation. The novelty of the study lies in combining a speckle imaging subpixel correlation algorithm with ML classification and with pre- and post-processing. This approach enables early automated assessment of antimicrobial effects with potential applications in rapid drug susceptibility testing and microbiological research.