RESUMEN
Here we describe embGAN, a deep learning pipeline that addresses the challenge of automated cell detection and tracking in label-free 3D time lapse imaging. embGAN requires no manual data annotation for training, learns robust detections that exhibits a high degree of scale invariance and generalizes well to images acquired in multiple labs on multiple instruments.
RESUMEN
Patterns of lineal descent play a critical role in the development of metazoan embryos. In eutelic organisms that generate a fixed number of somatic cells, invariance in the topology of their cell lineage provides a powerful opportunity to interrogate developmental events with empirical repeatability across individuals. Studies of embryonic development using the nematode Caenorhabditis elegans have been drivers of discovery. These studies have depended heavily on high-throughput lineage tracing enabled by 4D fluorescence microscopy and robust computer vision pipelines. For a range of applications, computer-aided yet manual lineage tracing using 4D label-free microscopy remains an essential tool. Deep learning approaches to cell detection and tracking in fluorescence microscopy have advanced significantly in recent years, yet solutions for automating cell detection and tracking in 3D label-free imaging of dense tissues and embryos remain inaccessible. Here, we describe embGAN, a deep learning pipeline that addresses the challenge of automated cell detection and tracking in label-free 3D time-lapse imaging. embGAN requires no manual data annotation for training, learns robust detections that exhibits a high degree of scale invariance, and generalizes well to images acquired in multiple labs on multiple instruments. We characterize embGAN's performance using lineage tracing in the C. elegans embryo as a benchmark. embGAN achieves near-state-of-the-art performance in cell detection and tracking, enabling high-throughput studies of cell lineage without the need for fluorescent reporters or transgenics.