Building a three-dimensional model of early-stage zebrafish embryo brain.

We introduce a computational approach to build three-dimensional (3D) surface mesh models of the early-stage zebrafish brain primordia from time-series microscopy images. The complexity of the early-stage brain primordia and lack of recognizable landmarks pose a distinct challenge for feature segmentation and 3D modeling. Additional difficulty arises because of noise and variations in pixel intensity. We overcome these by using a hierarchical approach in which simple geometric elements, such as "beads" and "bonds," are assigned to represent local features and their connectivity is used to smoothen the surface while retaining high-curvature regions. We apply our method to build models of two zebrafish embryo phenotypes at discrete time points between 19 and 28 h post-fertilization and collect measurements to quantify development. Our approach is fast and applicable to building models of other biological systems, as demonstrated by models from magnetic resonance images of the human fetal brain. The source code, input scripts, sample image files, and generated outputs are publicly available on GitHub.