Estimating the growth rate of microbes can give insights into how growth changes in times of stress, and more importantly, these findings can have implications for studying the effects of antibiotics. Computing single-cell growth rate — which is reported as volume change — is however computationally challenging since it includes determining cell outlines, extrapolating volumes, assigning buds to mothers, and identifying budding events. Notably, cell outlines overlap with each other, which makes it hard to determine cell shape, and extrapolating volumes is difficult since approximating cell shapes to be spherical or rod-shaped leads to a loss of information regarding the properties of the cells. In addition, buds — especially smaller ones — usually overlap with their mothers or neighbors, making it difficult to identify mother-and-bud pairs; buds are also notoriously hard to be detected by eye. In order to address these challenges, in a recent study, Peter S. Swain and colleagues developed a framework called Birth Annotator for Budding Yeast, or BABY, to determine single-cell growth rates from label-free images of budding yeast.
BABY uses a U-net — which is a convolutional neural network — and the main advantage of the method lies in the selection of training images that are used. The authors observed that cell size affects cell overlap, with most overlaps occurring between mid-sized cells and buds. To reduce overlaps between cells within each category in the imaging data, three training images were generated from each annotated image, each of them showing cells in the small, medium, and large categories, respectively. To elucidate growth rates and overlaps, the authors defined the training target as the overlap between any pair of cells and the boundary between the mother and bud in budding microorganisms, meaning, the ‘bud neck.’ The formation of a bud neck is sometimes visible as a darkening of the buds and this can help identify which bud belongs to which cell.
This is a preview of subscription content, access via your institution