Skeletonization TASK-03
Creating skeletal representations of neuron morphology for analysis.
Task Overview
What is Skeletonization?
Skeletonization transforms dense 3D neuron segmentations into simplified skeletal representations—graphs of connected nodes that trace the centerline of each process. These skeletons capture the essential branching structure while reducing data complexity by orders of magnitude.
Why is it Important?
Skeletons enable quantitative morphological analysis: branch length, tortuosity, branching patterns, and cable properties. They're essential for comparing neuron types, building computational models, and visualizing connectivity. Most downstream analysis tools work with skeletons rather than dense voxel data.
Key Objectives
- Generate accurate centerline traces through neuron volumes
- Correctly identify and mark branch points
- Preserve morphological features (spine locations, varicosities)
- Validate skeleton connectivity matches segmentation
- Export in standard formats (SWC, JSON)
Current Standard Operating Procedure Authoritative
Scope
This SOP covers both automated skeletonization validation and manual skeleton correction. Most skeletons are generated automatically; human review focuses on quality control and fixing errors.
Automated Skeleton Review
- Load the neuron and its generated skeleton in Neuroglancer
- Enable skeleton overlay on the 3D view
- Verify skeleton runs through center of each process (not along edges)
- Check all branch points are correctly detected
- Verify terminal nodes reach actual process endings
- Flag errors for manual correction
Manual Skeleton Correction
- Open skeleton editor in WebKnossos
- Navigate to flagged error location
- For missing branches: add nodes tracing the centerline
- For incorrect branch points: split and reconnect nodes
- For centerline drift: adjust node positions
- Save and re-validate
Node Placement Guidelines
- Place nodes at morphological landmarks (branch points, diameter changes)
- Space intermediate nodes every 1-2 µm along straight segments
- Node should be at volumetric center, not membrane
- Maintain smooth interpolation between nodes
Quality Criteria
- Skeleton fully connected (no orphan branches)
- Centerline deviation less than 0.5 µm from true center
- All branch points within 1 node of actual bifurcation
- No skeleton paths through extracellular space
Historical SOPs Archived
Previous skeletonization protocols are preserved here for reference. Always use the Current SOP above.
Version History
-
v1.0 (Current) — 2025
Initial automated + manual hybrid workflow
Visual Examples
Reference images showing correct skeleton placement and common errors.
Skeleton Quality Reference
- Good centerline — Skeleton runs through middle of process
- Edge-riding (error) — Skeleton hugs one side of membrane
- Missing branch — Process exists but no skeleton nodes trace it
- Floating segment — Skeleton disconnected from main tree
Common Failure Modes
- Missing thin processes Very fine axons may be skipped by automated skeletonization. Manually trace any visible processes without skeleton coverage.
- Incorrect branch point topology Branch points should have exactly one parent and multiple children. Fix loops or disconnections immediately.
- Centerline drift in curved regions Automated skeletons may cut corners. Add intermediate nodes to follow actual curvature of the process.
- Skeleton extending beyond segmentation Nodes should never be in extracellular space. Delete or reposition nodes that extend past the neuron boundary.
- Inconsistent node density Maintain roughly uniform spacing. Dense clusters make analysis harder; sparse regions lose morphological detail.
Tools Used
WebKnossos
Skeleton editing and manual tracing
Neuroglancer
3D skeleton visualization and validation
Kimimaro
Automated skeleton generation library
TEASAR
Tree-structure extraction algorithm
Output Formats
- SWC — Standard format for neuron morphologies
- JSON — Structured format with metadata
- Precomputed — Neuroglancer-compatible chunks
Training Videos
Related Publications
-
TEASAR: Tree-structure Extraction Algorithm for Accurate and Robust Skeletons