NA-MIC Project Weeks

Back to Projects List

SlicerAdaptiveBrush - Adaptive Brush Segment Editor Effect

Key Investigators

• Ben Zwick (The University of Western Australia and Talk2View, Australia)
• Andy Huynh (Talk2View, Australia)

Project Description

SlicerAdaptiveBrush is a segment editor effect extension for 3D Slicer that provides an adaptive brush tool for semi-automatic segmentation. The brush automatically segments regions based on image intensity similarity within the brush area, adapting to image features (edges, boundaries) rather than using a fixed geometric shape.

Objective

1. Submit to Extension Index
2. Improve documentation and tutorials
3. Optimize performance for real-time interaction
4. Add GPU acceleration for Level Set algorithm

Approach and Plan

1. Submit to Extension Index

• Complete submission requirements
• Test on all platforms (Linux, macOS, Windows)
• Create extension icon and screenshots

2. Improve documentation

• Write user tutorial with example workflows
• Document algorithm selection guide
• Add parameter tuning recommendations

3. Optimize performance

• Profile and optimize critical paths
• Implement ROI result caching for nearby brush positions
• Add slice-by-slice preview mode

4. GPU acceleration

• Implement OpenCL/CUDA backend for Level Set
• Benchmark CPU vs GPU performance

Progress and Next Steps

Completed

1. Extension Index Submission (In Progress)
   • CI/CD pipeline with GitHub Actions for automated builds
   • Extension not yet in the Slicer Extension Index
2. Documentation
   • Full documentation site live at benzwick.github.io/SlicerAdaptiveBrush
   • Getting Started Tutorial with 10-step workflow
   • Algorithms Guide covering all 7 algorithms
   • Parameter Wizard Guide for interactive setup
   • Developer documentation for optimization, testing, and recipes
   • Auto-generated screenshots from test suite
3. Performance Optimization
   • Implemented PerformanceCache with gradient and threshold caching
   • Undo/redo integration with single save per stroke
   • Cache statistics and hit rate logging
4. GPU Acceleration
   • Backend selector UI prepared (Auto/CPU/GPU)
   • GPU implementation deferred to v2.0+

Next Steps

• Complete Extension Index submission
• Parameter optimization and testing for different image modalities (CT, MRI T1/T2, PET) and tissue types (tumor, bone, vessels, brain tissue)
• Testing with Imaging Data Commons data using Claude skills:
  • ImagingDataCommons/idc-claude-skill
  • mhalle/idc-skill
  • benzwick/imaging-data-commons-skill
  • See also: claude-scientific-skill for Imaging Data Commons project
• Mouse shortcuts can be configured using SlicerMouseMaster for workflow optimization

Illustrations

Selecting the Adaptive Brush effect in Segment Editor

Painting a brain tumor segmentation - the brush adapts to image boundaries

3D surface rendering of the segmented tumor

Background and References

Code repository: https://github.com/benzwick/SlicerAdaptiveBrush

Documentation: https://benzwick.github.io/SlicerAdaptiveBrush/

• Getting Started Tutorial
• Algorithms Guide
• Parameter Wizard

Features

• Multiple algorithm choices - Geodesic Distance, Watershed, Random Walker, Level Set, Connected Threshold, Region Growing, Threshold Brush
• Auto-threshold methods - Otsu, Huang, Triangle, Maximum Entropy, IsoData, Li
• Automatic intensity analysis - GMM-based threshold estimation adapts to image content
• Edge-aware boundaries - Respects anatomical boundaries automatically
• 2D and 3D modes - Works on single slices or volumetrically (sphere mode)

References

• ITK-SNAP Adaptive Brush - Original inspiration
• 3D Slicer Segment Editor
• SlicerSegmentEditorExtraEffects