Related Experiment Video
Updated: Jun 16, 2026
10:41
Tracking and Quantifying Developmental Processes in C. elegans Using Open-source Tools
Published on: December 16, 2015
Using machine learning to speed up manual image annotation: application to a 3D imaging protocol for measuring single
Zafer Aydin1, John I Murray, Robert H Waterston
1Department of Genome Sciences, University of Washington, Seattle, WA 98195, USA. zafer@u.washington.edu
BMC Bioinformatics
|February 12, 2010
Summary
This study developed a machine learning approach to automate error correction in StarryNite, a cell tracking program for C. elegans development. The new method significantly reduces the time needed for manual error annotation in biological image analysis.
Area of Science:
- Developmental Biology
- Computational Biology
- Machine Learning
Background:
- Image analysis is crucial for studying gene expression, cell cycle, and protein localization in biological experiments.
- 3-D time lapse microscopy and the StarryNite program are used for tracking C. elegans gene expression and cell lineage.
- Manual error correction of StarryNite's output using AceTree is time-consuming, taking several hours per experiment.
Purpose of the Study:
- To reduce the time required for manual error correction in StarryNite.
- To develop an automated method for identifying and correcting frequent errors in cell division calls made by StarryNite.
Main Methods:
- A support vector machine (SVM) classifier was trained to distinguish correct cell divisions from misclassified movements.
- Cross-validation experiments were performed on benchmark datasets to evaluate the SVM's performance.
Main Results:
- The trained SVM classifier significantly and successfully identifies movement errors misclassified as divisions.
- The developed SVM classifier has been integrated into a new version of the StarryNite program.
Conclusions:
- Machine learning offers a viable approach for automating error annotation in biological image analysis tools like StarryNite.
- The study provides general methodologies for developing and validating classifiers for pattern recognition tasks in biological data.
