NUCLEAR SEGMENTATION IN MICROSCOPE CELL IMAGES: A HAND-SEGMENTED DATASET AND COMPARISON OF ALGORITHMS

Area of Science:

• Microscopy and image analysis
• Computational biology
• Cell biology

Background:

• Image segmentation is crucial for image analysis but often lacks objective evaluation.
• Existing methods are frequently assessed subjectively or with limited data.
• High-throughput biological studies require robust and minimally interactive segmentation tools.

Purpose of the Study:

• To address the lack of objective benchmarks for image segmentation algorithms.
• To evaluate existing segmentation methods in a high-throughput context.
• To provide a publicly available dataset and software for reproducible research.

Main Methods:

• Hand-segmentation of 97 fluorescence microscopy images, totaling 4009 cells.
• Objective evaluation of previously proposed image segmentation algorithms.
• Focus on algorithms suitable for high-throughput settings with minimal user intervention.

Main Results:

• Established a comprehensive, hand-labeled dataset for fluorescence microscopy image segmentation.
• Provided an objective framework for comparing segmentation algorithm performance.
• Demonstrated the utility of the dataset for evaluating algorithms in high-throughput applications.

Conclusions:

• The developed dataset and evaluation framework enable objective benchmarking of image segmentation algorithms.
• This resource facilitates the advancement of automated cell segmentation in high-throughput microscopy.
• Public availability promotes reproducibility and standardization in biological image analysis.