Active mesh and neural network pipeline for cell aggregate segmentation.

Area of Science:

• Cell Biology
• Biotechnology
• Microscopy

Background:

• 3D cell segmentation in cellular aggregates is increasingly challenging due to advanced microscopy.
• Accurate cell segmentation and tracking are crucial for understanding cellular dynamics.

Purpose of the Study:

• To develop and validate a novel pipeline for segmenting and tracking cells within 3D cellular aggregates.
• To integrate neural network predictions with active mesh deformation for robust segmentation.

Main Methods:

• A pipeline combining neural network segmentation with active mesh deformation was developed.
• The method was applied to mouse mammary gland organoids and mouse embryonic stem cells.
• A Fiji plugin was created for user interaction, training data generation, and segmentation correction.

Main Results:

• The pipeline successfully segmented individual cells using nuclear and membrane markers.
• Cell tracking over time was demonstrated in 3D organoid and stem cell cultures.
• Metrics for quantifying automated segmentation quality were established.

Conclusions:

• The active meshes-based approach offers efficient postprocessing and correction for 3D cell segmentation.
• This pipeline enhances the integration of neural network predictions with experimental microscopy data.
• The method is versatile and applicable to various cell types and microscopy techniques.