Deep Learning-Based Instance Segmentation and Lineage Tracking of Clonal Organisms

Department

Nanoengineering, Joint School of Nanoscience and Nanoengineering

Document Type

Poster

Publication Date

4-17-2026

Abstract

Time-lapse microscopy generates large imaging datasets that enable detailed study of how organisms grow and divide at the individual level. For clonal organisms that reproduce by budding (yeast) or fission (bacteria), resolving individual growth trajectories and reconstructing mother-daughter lineage relationships are essential for developmental, physiological, and biotechnological studies. Deep learning pipelines such as DeLTA have demonstrated that automated segmentation, tracking, and lineage reconstruction are achievable for bacteria in both microfluidic devices (controlled) and on agarose pads (uncontrolled), but this capability has not been extended to clonal plant systems, where organisms (i.e., entire individuals) are larger, morphologically variable, and overlap in dense populations. Existing plant phenotyping platforms measure only population-level metrics such as total area or colony count without resolving individual lineage structure. We present an open-source framework that combines U-Net-based instance segmentation with Bayesian multi-object tracking to automatically identify, track, and reconstruct clonal lineages from time-lapse microscopy data. To validate the framework, we use duckweed, the smallest flowering plant. We analyzed individuals grown in standard petri dishes. The pipeline tracked dozens of unique plants for about a week, identified budding events spanning two generations, and measured the average doubling times for populations. Segmentation achieved a Dice coefficient of 0.94 (petri dish), with full analysis completing in approximately 30 minutes on a GPU workstation. Future work will extend the analysis to plants grown in microfluidic chips. This framework extends deep learning-based lineage analysis beyond microbial systems to clonal plant biology and has potential applicability to other clonally reproducing organisms.

This document is currently not available here.

Share

COinS