Monte Carlo simulation of cone-beam CT (CBCT) projections using OpenGATE 10.x.

This project simulates a dental CBCT scan by:

1. Creating an acrylic cylinder phantom with embedded ball bearing markers (BBs) arranged in a precise 3-plane pattern
2. Rotating the phantom through 360° while keeping source and detector fixed
3. Generating projection images at each angle using Monte Carlo particle transport
4. Providing interactive visualization tools to analyze the projection data

Key Features:

• Realistic physics simulation using Geant4 via OpenGATE
• Customizable phantom geometry and materials
• Multiple visualization modes (2D/3D, interactive/static)
• Automatic saving of all plots for publication
• Jupyter notebook interface for interactive exploration

# Clone or download this repository
cd cbct/

# Install dependencies
pip install -r requirements.txt

python simulation.py

This will:

• Generate 12 projections (default, adjustable in code)
• Save individual projections and combined stack to output/
• Typical runtime: varies with particles_per_proj (default: 5 billion photons/projection)

Output files:

output/
├── projections_counts.mhd     # Combined projection stack (1504 × 1248, uint16)
├── projections_counts.raw     # Binary data
├── proj_single_counts.mhd     # Last single projection
└── *.root                     # ROOT files (hits, singles)

Command-line interactive menu:

python visualize.py

Select from 9 visualization modes:

• 1-4: 2D views (slice browser, montage, sinogram, statistics)
• 5-9: 3D views (orthogonal slices, surface plots, volume rendering)

Jupyter notebook (interactive slider):

jupyter notebook slice_browser.ipynb

All plots are auto-saved to output/plots/ as PNG (600 dpi) or HTML (Plotly).

cbct/
├── simulation.py          # Monte Carlo CBCT simulation (OpenGATE)
├── visualize.py           # Visualization suite (matplotlib + plotly)
├── slice_browser.ipynb    # Interactive Jupyter notebook
├── requirements.txt       # Python dependencies
├── README.md              # This file
└── output/                # Generated data and plots
    ├── projections_counts.mhd
    └── plots/             # Auto-saved visualizations
        ├── projection_montage.png
        ├── sinogram.png
        ├── 3d_isosurface.html
        └── ...

Adjust these in simulation.py (lines 18-27):

Detector & Hardware:

• Detector: Mercu1012X CsI flat-panel (300.8 × 249.6 mm, 100 µm native pixels)
• Source: RADII KL3SB-0.5-130 / Gemini012 gantry (0.5 mm focal spot)
• Matrix mode: Binning 2×2 → 1504 × 1248 @ 0.2 mm spacing
• CsI thickness: 700 µm scintillator

Phantom anatomy:

• Acrylic cylinder (57–60 mm inner/outer radius, 27.5 mm height)
• 27 regular ball bearings (1.5 mm dia., stainless steel) arranged in 3 horizontal planes at y = 22.5, −2.5, −22.5 mm
  • Top plane: 12 BBs at 30° intervals; 1 marker at θ=0°
  • Middle plane: 12 BBs at 30° intervals; 1 marker at θ=180°
  • Bottom plane: 6 BBs at 60° intervals; 1 marker at θ=0°
• 3 marker ball bearings (2.5 mm dia., stainless steel) at specific angles for geometry verification

Physics: G4EmStandardPhysics_option4 with 1 mm production cuts

• Engine: Geant4 10.x via OpenGATE
• Particle transport: Photons (gamma) with full EM physics
• Geometry: Constructive solid geometry (CSG) volumes
• Materials: NIST database (G4_WATER, G4_BONE_CORTICAL_ICRP, etc.)
• Detector: CsI scintillator, energy-weighted digitizer

• Image format: MetaImage (.mhd/.raw) - universally readable
• Array shape: (n_proj, ny, nx) = (12, 1248, 1504) with default settings
• Data type: uint16 (16-bit unsigned, 0–65535 counts, matching detector ADC)
• Pixel spacing: 0.2 mm (200 µm) native / 0.4 mm (400 µm) binned
• Value units: Photon counts per pixel (energy-weighted digitizer)

• Z-axis: Source-detector axis (source at −480 mm, isocenter at 0, detector at +220 mm)
• Y-axis: Vertical axis (phantom cylinder axis after tilting)
• Rotation: Phantom rotates around Y-axis per projection angle
• Detector: Fixed 2D flat panel perpendicular to Z-axis