Reproducibility assessment of a published cell shape transition analysis workflow.

Area of Science:

• Biomedical research
• Image analysis
• Cell biology

Background:

• Reproducibility in image-based studies is a significant challenge.
• Limited access to raw data and incomplete analysis workflow reporting contribute to this issue.
• The Global Bioimage Analyst Society (GloBIAS) initiated an effort to assess published bioimage analysis workflows.

Purpose of the Study:

• To evaluate the reproducibility of a published bioimage analysis workflow for Dictyostelium discoideum cell shape transitions.
• To identify specific steps within the workflow that impede reproducibility.
• To determine if improved segmentation quality enhances the reproducibility of downstream analyses.

Main Methods:

• Evaluation of a published cell shape transition analysis workflow.
• Attempted reproduction of cell segmentation using reported parameters.
• Application of an alternative segmentation strategy using ilastik, a machine learning tool.
• Reconstruction of kymographs using cell-boundary coordinates from alternative segmentation.

Main Results:

• Published kymographs and biological conclusions were qualitatively reproducible.
• The cell segmentation step was insufficiently described, requiring extensive parameter tuning and yielding inaccurate cell contours.
• Utilizing ilastik for improved segmentation allowed for partial reconstruction of kymographs, partially matching published results.
• Accessible software and raw data alone do not guarantee reproducibility.

Conclusions:

• Complete reporting of image processing workflows, including segmentation details and parameter settings, is essential for reproducible bioimage analysis.
• Incomplete reporting of critical steps like cell segmentation compromises downstream analysis and overall study reproducibility.
• Enhancing segmentation quality can partially improve the reproducibility of downstream quantitative analyses.