Abstract
This protocol is intended for use on light microscopy datasets of larval zebrafish ubiquitously expressing soma- and membrane-based fluorescent proteins. Methods are subject to change based on ultimate segmentation goals and expansion into variable light microscopy datasets. Key components include Amira project setup, mask segmentation, and an overview of class types and segmentation techniques.
Introduction
Zebrafish are a cornerstone of developmental biology, but the transition from raw 3D imaging to quantitative anatomical analysis remains a significant bottleneck. When using ubiquitous fluorescent markers—where nearly every cell or membrane is labeled—the resulting data often appears as a complex "soup" of signal. Distinguishing specific muscle bundles from adjacent organs requires more than spatial knowledge; it necessitates recognizing subtle differences in signal intensity, texture, and boundary patterns.
This protocol provides a comprehensive roadmap for segmenting larval zebrafish organs and substructures from 3D light microscopy data using the Amira 3D Classic Segmentation Workroom. By defining specific visual textures and anatomical landmarks, it offers a standardized workflow for transforming complex, ubiquitously labeled fluorescent datasets into high-fidelity 3D anatomical models.
Methods
The protocol utilizes Amira 3D 2022.2 and focuses on a manual-to-semi-automated segmentation pipeline designed for light microscopy datasets of larval zebrafish.
Preprocessing and Setup
The workflow begins by importing raw TIFF data and resampling voxel sizes to half the original resolution to improve label resolution. An initial "fish mask" is created to define global sample boundaries, which aids in later mesh visualization and maintains anatomical context.
Segmentation Strategy
Labels are organized into distinct "label fields" (e.g., Muscle, Cartilage, Organs) to prevent accidental overwriting and manage data complexity. The protocol suggests segmenting each class in its "best" fluorescent channel and overlaying results post-export.
Refinement Techniques
- Interpolation: Users annotate intermittent slices in the XY plane and use interpolation (
Ctrl+I) to fill gaps. - Smoothing: Global smoothing is applied, though the authors caution that excessive smoothing can cause "material shrinking," particularly in datasets with limited Z-dimension depth.
Results and Key Insights
The primary output of this protocol is a detailed atlas of visual cues for major systems in the larval zebrafish.
Texture-Based Identification
The methodology moves beyond simple "blob finding" by using texture as a diagnostic tool. For instance, cartilage is identified by a "hollow, porous, and bright" pattern, while the liver is recognized by a unique "gradient of intensity." In regions of overlapping signals, a hierarchy is established: muscle fibers and cartilage are segmented first as they provide the most reliable structural boundaries.
Anatomical Coverage
The workflow enables the detailed identification of musculature (trunk, pectoral girdle, abdominal, and cranial), skeletal structures (neurocranium, cleithrum, and notochord), internal organs (heart, kidney nephrons, swim bladder, and GI tract), and the nervous system.
The resulting 3D reconstructions accurately represent the spatial relationships of these complex systems, providing a high-fidelity model for further analysis.
Discussion and Limitations
While the protocol provides a robust framework, the process remains highly manual and time-consuming, requiring significant anatomical expertise. Identifying boundaries based on "soft gray dots" or "webbed signal patterns" introduces user subjectivity that may affect reproducibility across different laboratories.
Future work aims to use these high-quality manual segmentations as "ground truth" training data for machine learning algorithms. This transition toward automation will be essential for scaling anatomical analysis in zebrafish research.
Summary updated