Optimized U-Net model for 3D light-sheet image segmentation of zebrafish trunk vessels

Area of Science:

• Developmental Biology
• Bioimaging
• Computational Biology

Background:

• Zebrafish vasculature serves as a model for studying vascular development and biological mechanisms.
• High-throughput light-sheet microscopy generates large 3D datasets of zebrafish trunk vessels.
• Automated and quantitative analysis of these complex datasets is currently limited by segmentation accuracy.

Purpose of the Study:

• To develop an accurate and automated method for segmenting 3D zebrafish trunk vasculature.
• To enable quantitative analysis of zebrafish vascular growth.
• To address the bottleneck in analyzing large-scale 3D vascular imaging data.

Main Methods:

• Proposed a novel Multi-scale 3D U-Net model for semantic segmentation.
• Applied the model to segment intersegmental vessels and dorsal longitudinal anastomotic vessels in zebrafish trunk.
• Utilized light-sheet fluorescent microscopy for high-resolution 3D imaging.

Main Results:

• Achieved high segmentation accuracy for intersegmental vessels (82.3% IoU).
• Achieved high segmentation accuracy for dorsal longitudinal anastomotic vessels (83.0% IoU).
• Enabled quantitative analysis of zebrafish vasculature growth between 42-62 hours post-fertilization.

Conclusions:

• The Multi-scale 3D U-Net model provides an effective solution for accurate zebrafish trunk vessel segmentation.
• This automated approach facilitates quantitative analysis of vascular development.
• The method advances the study of biological mechanisms underlying vascular growth using zebrafish models.