3DCellComposer - A versatile pipeline utilizing 2D cell segmentation methods for 3D cell segmentation.

Area of Science:

• Bioimage Informatics
• Computational Biology
• Cellular Imaging

Background:

• Accurate cell segmentation is vital for cellular analyses in bioimage informatics.
• Existing 3D cell segmentation methods struggle with limited training data and 3D complexity, lacking broad applicability.
• Current methods often require extensive human annotation for new imaging modalities.

Purpose of the Study:

• To develop a generalizable method for accurate 3D cell segmentation using existing 2D segmentation models.
• To create an open-source package, 3DCellComposer, that integrates 2D segmentation models without retraining.
• To enhance the CellSegmentationEvaluator tool for 3D image quality assessment without human annotations.

Main Methods:

• Implemented a novel approach to adapt 2D cell segmentation models for 3D image analysis.
• Developed the versatile, open-source software package 3DCellComposer.
• Enhanced the CellSegmentationEvaluator to provide quantitative metrics for 3D image segmentation performance.

Main Results:

• 3DCellComposer successfully produced high-quality 3D segmentations of multichannel tissue images.
• The approach demonstrated broad applicability across different imaging modalities by leveraging existing 2D models.
• The enhanced CellSegmentationEvaluator enabled objective performance assessment without human annotations.

Conclusions:

• 3DCellComposer offers a valuable alternative to training new 3D segmentation models, especially when human-annotated 3D data is scarce.
• This method significantly reduces the effort required for 3D cell segmentation across diverse biological samples and imaging techniques.
• The tool is poised to benefit large-scale bioimaging initiatives like the Human BioMolecular Atlas Program.