How hutin1/aso works
Project Type and Purpose
• This is a 3D Slicer extension (plugin) called Automated Standardized Orientation (ASO).
• It provides a GUI for orienting 3D scans (CBCT and IOS) in two modes:
• Semi-Automated – landmark-based ICP registration only
• Fully-Automated – includes Pre-Orientation (deep learning), Landmark Identification (ALI_CBCT), then ICP
File-Structure Analysis
Top-level:
• ASO/ – core Slicer extension
– ASO.py (entry point/controller)
– Methode/ (generic pipeline classes: Methode.py, CBCT.py, IOS.py, Progress.py)
– Testing/ (CMake test harness, Python tests)
– CMakeLists.txt – build config for Slicer extension
• ASO_CBCT/ – CBCT-specific pipelines
– PRE_ASO_CBCT/ & SEMI_ASO_CBCT/ (mode implementations: .py + .xml descriptors)
– utils/ (NeuralNet loader, resampling, common routines)
• ASO_IOS/ – IOS-specific pipelines
– PRE_ASO_IOS/ & SEMI_ASO_IOS/ (mode implementations)
– utils/ (cache, ICP, pre_icp, transformations, I/O helpers)
• CMakeLists.txt & LICENSE.txt & README.md
README Insights
• Describes two modes with their inputs/outputs
• Highlights external dependencies: deep-learning PreASO models, ALI_CBCT for landmark identification, SlicerDentalModelSeg for segmentation
• Shows high-level flowcharts for CBCT and IOS pipelines
Key Components
a. User Interface
• 3D Slicer GUI extension (ASO.ui)
• Controls: Input type (CBCT/IOS), Mode selector, Folder pickers, Run button, Output options
b. Controller Layer
• ASO.py orchestrates user input → pipeline dispatch
• Methode.py defines abstract pipeline interface, Progress.py updates GUI
c. Processing Pipelines
• CBCT Pipeline
• Pre-Orientation (deep learning model loader & predictor – utils/Net.py)
• Landmark Identification (ALI_CBCT external module)
• ICP Registration (Methode/CBCT.py + ASO_CBCT utils)
• IOS Pipeline
• (Optional) Pre-ICP cleaning steps (utils/pre_icp.py)
• ICP Registration (Methode/IOS.py + ASO_IOS utils/icp.py)
d. Data Management
• Inputs: Scan files (.nii, .vtk…), Landmark files, Reference gold files, Models folders
• Outputs: Oriented meshes/volumes, logs
e. External Services & Dependencies
• 3D Slicer host application
• Deep-learning models (downloaded .zip)
• ALI_CBCT library (landmark detection)
• SlicerDentalModelSeg (for IOS segmentation)
Architectural Patterns & Principles
• Plugin Architecture: Clear separation between host (3D Slicer) and extension code
• Modular Pipeline: Abstract base class for pipelines (Methode) with CBCT/IOS specializations
• Separation of Concerns:
• UI vs. Orchestration vs. Processing vs. Data I/O
• Fully vs. Semi modes share common ICP step, differ in preprocessing