A Deep Learning Pipeline for Cell Segmentation and Viability Quantification in 3D Constructs From Fluorescence...

Area of Science:

• Biomedical Engineering
• Cell Biology
• Microscopy Imaging

Background:

• 3D cell cultures and bioprinting are increasingly used in research.
• Fluorescence microscopy in 3D faces challenges like signal attenuation and out-of-focus interference, hindering automated analysis.

Purpose of the Study:

• To develop an automated tool for cell segmentation and viability assessment in 3D fluorescence microscopy images.
• To overcome limitations of traditional methods in 3D cell culture monitoring.

Main Methods:

• A hybrid pipeline combining deep learning (U²-Net) for segmentation and traditional image processing (watershed, intensity classification) for viability.
• Training on 2D cell cultures and application to 3D bioprinted human umbilical vein endothelial cells in gelatin methacrylate.

Main Results:

• The model achieved accurate cell segmentation with minimal overfitting.
• Quantitative comparisons confirmed accuracy against manual counts and ImageJ.
• Automated tracking of cell density and viability in 3D constructs closely matched manual assessments.

Conclusions:

• The developed hybrid pipeline provides a scalable, reproducible, and efficient automated method for analyzing 3D fluorescence microscopy data.
• This approach enables non-invasive, high-throughput monitoring of cell density and viability in 3D bioprinted constructs.