DeLTA: Automated cell segmentation, tracking, and lineage reconstruction using deep learning
Problem
Analyzing time-lapse microscopy movies of E. coli cells in a mother machine required slow, human-supervised segmentation, tracking and lineage reconstruction, limiting accuracy and throughput.
Why it matters
Manual curation of single-cell microscopy data bottlenecks biological research that depends on long-term, high-throughput cell analysis.
What the AI did
A deep learning pipeline automatically performed cell segmentation, tracking and lineage reconstruction without human supervision.
Results
1% error rate; analysis of a typical frame of ~150 cells in <700 msec
Impact
Enables unsupervised, high-fidelity, high-throughput analysis of mother-machine movies.
Record
| Venue | PLoS Computational Biology |
|---|---|
| Published | 2020 |
| Work type | article |
| Authors | Jean‐Baptiste Lugagne, Haonan Lin, Mary J. Dunlop |
| Citations | 242 (OpenAlex, at import time) |
| References | 41 |
| Model or method | deep learning (specific architecture not stated in source) |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1371/journal.pcbi.1007673 |
| Open access | gold |
| Keywords | deep learning, cell segmentation, tracking, lineage reconstruction, mother machine, E. coli |
Importance rating
real but narrower advance
Abstract calls microscopy image analysis a major bottleneck and the mother machine widely used for long-term single-cell analysis; scope is single-cell imaging research, no scale stated.