Benchmark for multi-cellular segmentation of bright field microscopy images

Area of Science:

• Biomedical Imaging
• Computational Biology
• Cell Biology

Background:

• Accurate multi-cellular segmentation of bright field microscopy images is crucial for quantifying in vitro cell migration.
• Existing segmentation tools lack a standardized public benchmark for performance evaluation and comparison.

Purpose of the Study:

• To establish a uniform framework and publicly available benchmark for evaluating multi-cellular segmentation algorithms in bright field microscopy images.
• To facilitate fair comparison and assessment of different segmentation tools.

Main Methods:

• A comprehensive dataset of 171 manually segmented bright field microscopy images was curated from diverse sources.
• The dataset was partitioned into 8 distinct subsets for rigorous evaluation.
• Three leading multi-cellular segmentation tools were assessed using a uniform benchmarking framework.

Main Results:

• The study presents the first public annotated dataset specifically designed for benchmarking multi-cellular segmentation in bright field microscopy.
• The benchmark enables objective evaluation and comparison of segmentation algorithm performance.

Conclusions:

• The developed benchmark resource provides a vital tool for the scientific community to assess segmentation algorithms.
• This annotated dataset supports fair evaluations and comparisons of current and future segmentation methods.
• Researchers are encouraged to utilize and contribute to this benchmark for advancing cell migration analysis.