Tracking bacteria at high density with FAST, the Feature-Assisted Segmenter/Tracker

Area of Science:

• Microbiology
• Biofilm research
• Microbial community dynamics

Background:

• Bacteria commonly form surface-attached communities called biofilms.
• Understanding individual cell behavior in biofilms is crucial but challenging.
• Existing microbial tracking software requires extensive manual parameter tuning or deep learning model training.

Purpose of the Study:

• To develop an automated and user-friendly software for tracking microbial cells in biofilms.
• To overcome the limitations of manual parameter adjustment and data-intensive training in existing tracking methods.
• To enable high-throughput analysis of microbial behavior in complex communities.

Main Methods:

• Developed Feature-Assisted Segmenter/Tracker (FAST) using unsupervised machine learning and information theory.
• FAST quantifies unique information from distinguishing cell features to minimize tracking errors.
• Integrated segmentation, data visualization, lineage assignment, and manual track correction tools.

Main Results:

• FAST significantly reduces tracking errors compared to position-only methods (4-10 fold fewer errors).
• The unsupervised approach minimizes the need for manual parameter optimization and qualitative assessments.
• The modular design allows for extensibility and integration of custom image analyses.

Conclusions:

• FAST provides a powerful, efficient, and user-friendly solution for analyzing microbial cell behavior in biofilms.
• The software enables high-throughput, data-rich microbial community studies with minimal user intervention.
• FAST is available as a standalone application or in Matlab, with comprehensive documentation and tutorials.