Misic, a general deep learning-based method for the high-throughput cell segmentation of complex bacterial communities

Abstract

Studies of bacterial communities, biofilms and microbiomes, are multiplying due to their impact on health and ecology. Live imaging of microbial communities requires new tools for the robust identification of bacterial cells in dense and often inter-species populations, sometimes over very large scales. Here, we developed MiSiC, a general deep-learning-based 2D segmentation method that automatically segments single bacteria in complex images of interacting bacterial communities with very little parameter adjustment, independent of the microscopy settings and imaging modality. Using a bacterial predator-prey interaction model, we demonstrate that MiSiC enables the analysis of interspecies interactions, resolving processes at subcellular scales and discriminating between species in millimeter size datasets. The simple implementation of MiSiC and the relatively low need in computing power make its use broadly accessible to fields interested in bacterial interactions and cell biology.

Omnipose: a high-precision morphology-independent solution for bacterial cell segmentation

“…Given the goal of developing software with the capacity to recognize bacteria universally, we sought to identify strongly performing algorithms for further development. An unbiased, quantitative comparison of cell segmentation algorithms on bacterial cells has not been performed; thus, we selected one or more representatives from each category for our analysis: Morphometrics 23 ( i ), SuperSegger 13 ( ii ), Mask R-CNN 27 , StarDist 26 , MiSiC 15 , and Cellpose 12 ( iii ). For a detailed review of these choices, see Methods.…”

“…Given that SuperSegger was motivated at least in-part to mitigate these issues, we postulate that traditional image segmentation approaches are ultimately limited to specialized imaging scenarios. Although we classify MiSiC as a DNN-based approach, this algorithm also relies on thresholding and watershed segmentation to generate cell masks from its network output 15 . The network output of MiSiC is more uniform than unfiltered phase contrast images, yet this pre-processing does not fully abrogate the typical errors of thresholding and watershed segmentation.…”

“…It was also used in the segmentation and tracking package Usiigaci 24 . U-Net architectures have been implemented in a number of algorithms, including DeLTA, PlantSeg, MiSiC, StarDist, and Cellpose 12,15,17,22,26 . DeLTA was not included in this study because it operates similarly to MiSiC and was designed specifically for mother machine microfluidics analysis.…”

“…Trained DNNs are thus limited in applicability to images that are representative of those in the training dataset. Early DNN approaches were based on the Mask R-CNN architecture 24 , whereas more recent algorithms such as StarDist, Cellpose, and MiSiC are based on the U-Net architecture 12,15,26 . Pachitariu and colleagues showed that Cellpose outperforms Mask R-CNN and StarDist on a variety of cell types and cell-like objects, distinguishing it as a general solution for cell segmentation 12 .…”

SyMBac: Synthetic Micrographs for Accurate Segmentation of Bacterial Cells using Deep Neural Networks

“…Synthetic data has been used previously to aid in segmentation. MiSic [26] is a tool for segmenting bacterial micrographs based on real training data, but supplements this data with "synthetic" data, which however does not accurately reflect the optics, physics, or biology of the experiment which is being analysed, and thus cannot segment images based purely on synthetic data. BCM3D [27] is a tool used for fluorescence biofilm segmentation, however it relies on deconvolution, which can lead to increased noise and phantom objects in the output masks, or experimentally derived PSFs.…”