Digital holographic microscopy for bacterial species classification and motility characterization

Abstract

We have previously presented our work in developing and applying a commercial digital holographic microscopy (DHM) system for volumetric, 3D characterization of bacterial motility. The system was applied to simple biological systems, i.e., single bacterial species, to demonstrate its effectiveness. We are now applying DHM to more realistic conditions, including multiple bacterial types, to differentiate the species of interest for investigating their interactions. Our workflow for species classification and motility characterization combines DHM and machine learning. Specifically, our DHM instrument acquires holograms of single bacterial species and mixtures of different species, the software extracts in-focus images of individual bacteria and their trajectories, and deep convolutional neural network models are constructed and trained using the in-focus images and then deployed to the mixture data for classification. The motility and morphology of the predicted species in the mixture is consistent with the measurements from isolates, verifying the effectiveness of the developed workflow. This work showcases the application of DHM to investigate complex biological systems.