Three-dimensional reconstruction of live embryos using robotic macroscope images

Area of Science:

• Developmental Biology
• Biomedical Imaging
• Computer Vision

Background:

• Determining the three-dimensional (3-D) shape of live embryos is crucial for developmental biology research.
• Existing methods face challenges in accurately capturing dynamic embryonic morphology due to technical limitations.

Purpose of the Study:

• To develop and validate a novel 3-D reconstruction algorithm for live embryos.
• To address the technical challenges associated with imaging and reconstructing delicate biological specimens.

Main Methods:

• Utilizing a robotic macroscope to capture images of live embryos from multiple viewing angles.
• Employing block matching for establishing point correspondences between different image views.
• Implementing a new 3-D reconstruction algorithm that simultaneously estimates point positions and camera errors, assuming orthographic projections and known camera scaling factors.

Main Results:

• Demonstrated successful 3-D reconstructions of test objects and early-stage axolotl embryos.
• Showcased the necessity of accommodating camera positioning and alignment errors for meaningful reconstructions.
• Achieved a mean radius of error of 0.27 pixels (1.1 microns) in embryo surface reconstruction.

Conclusions:

• The developed algorithm effectively reconstructs the 3-D shape of live embryos with high accuracy.
• The ability to accommodate camera errors is critical for reliable 3-D morphological analysis of biological samples.
• This technique advances the field of live embryo imaging and analysis, providing a robust tool for developmental studies.