KAHIBARO
23.1. Introduction to Voxelized Geometry
Table of Contents
Medical images
In medical physics, voxelized geometry is the link between real patient images and Monte Carlo simulations. Instead of building a scene from simple boxes and cylinders, you import an image that already contains anatomical information and turn it into a 3D distribution of materials or activity. Each tiny box in this grid is a voxel, short for “volume pixel.”
Voxelized geometries usually come from medical imaging modalities. The most common inputs are CT, MRI, and nuclear medicine images, but they play different roles in simulation.
CT images are the primary source for building anatomical and material maps. Each CT slice is a 2D matrix of Hounsfield units, and when many slices are stacked you get a 3D image volume. This volume encodes how X rays were attenuated in the body and is used to infer electron density and material type, which are critical for dose and radiation transport simulations. In GATE you will later use CT information to derive material properties and assign them voxel by voxel.
MRI images provide excellent soft tissue contrast but do not directly give you electron density or Hounsfield units. In a voxelized geometry context, MRI can be used to define anatomical regions or to specify where certain structures lie, but an additional strategy is needed if you want to convert them into physical materials. For beginners, CT based voxelized geometry is much simpler and is usually the first choice.
Nuclear medicine images such as SPECT and PET do not describe anatomy or material. Instead, they describe how much activity is present in each voxel, that is, how many decays per second occur locally. In GATE, such images are commonly used to define voxelized sources where clinical activity distributions are reproduced in a simulation. This is central to internal dosimetry and radionuclide therapy studies and is conceptually different from CT based material maps, even though both are voxel grids.
The 3D nature of medical images is essential. Each image volume has a number of voxels along x, y, and z, a physical spacing between voxels in each direction, and an origin, which defines how the image is placed in space. In voxelized geometry, these pieces of information define the size and position of the voxel phantom in the simulation world. Later chapters will show how to read this metadata and map it to GATE geometry.
Finally, voxelized geometry is especially important for patient specific simulations. Instead of a generic water box, you can simulate dose in the actual patient anatomy or model scanner performance for a realistic phantom. This improves the clinical relevance of your results and is one of the major motivations for learning voxel-based approaches.
A voxelized geometry is built from a 3D medical image where each voxel has both a spatial size and a value. In dose simulations, CT based voxel values are converted to material and density, while in internal dosimetry and nuclear medicine, SPECT or PET voxel values define the spatial distribution of radioactive activity.
Voxel phantoms
A voxel phantom is the direct implementation of a medical image inside a Monte Carlo simulation. Conceptually, it is a 3D grid of tiny rectangular volumes, each with a defined position, size, and assigned property. In GATE, voxel phantoms provide a detailed representation of complex shapes that would be almost impossible to approximate accurately using a few simple geometric objects.
In an anatomical voxel phantom, each voxel is assigned a material and sometimes a density. These assignments usually come from CT based information and a predefined conversion scheme that classifies voxels into air, lung, soft tissue, bone, and other tissues. The result is a realistic digital body that can be placed in the world volume of a simulation. Radiation particles then travel through this heterogeneous structure and interact according to the local material in each voxel.
In a source voxel phantom, the main property attached to each voxel is radioactive activity. Here the values come from nuclear medicine images or synthetic activity maps. Each voxel may represent uniform activity or may store an activity density such as Bq per milliliter. GATE uses these values to sample decay locations so that particles are emitted according to the actual clinical or experimental distribution.
You can also combine the two concepts. A single voxel phantom can carry both material information and activity information, for instance by using one image for materials and another aligned image for activity. This is very common in internal dosimetry, where patient CT defines the anatomy and SPECT or PET defines where the radiopharmaceutical is located.
Voxel phantoms are characterized by their resolution and extent. The voxel size is determined by the image spacing. Smaller voxels provide finer anatomical detail but increase memory usage and computation time because more volumes must be tracked. Larger voxels reduce detail but are faster to simulate. When planning a voxelized simulation in GATE you need to balance realism and performance by choosing an appropriate resolution.
The physical size of the phantom is defined by the number of voxels and their spacing. For example, a CT with $512 \times 512 \times 200$ voxels and spacing of $0.5 \text{ mm} \times 0.5 \text{ mm} \times 1 \text{ mm}$ corresponds to a phantom that is $256 \text{ mm} \times 256 \text{ mm} \times 200 \text{ mm}$ in size. This determines how it fits inside the world volume and how close external sources or detectors must be placed.
Alignment and orientation are also key aspects of voxel phantoms. The phantom has its own coordinate system, which is determined by the image origin and direction. In GATE, you must ensure that this phantom coordinate system is correctly mapped to the global simulation coordinates. Misalignment leads to incorrect placement relative to beams, sources, or detectors and can invalidate results.
Voxel phantoms can represent either patients or standard computational models. Standardized computational phantoms are predefined voxelized anatomies that represent typical adult or pediatric bodies and are often used for regulatory or comparative studies. Patient specific phantoms are created from individual CT datasets and reflect the unique anatomy of a particular case. Both use the same underlying concept of voxelized geometry.
From the point of view of particle transport, a voxel phantom behaves like a densely packed collection of small volumes. Internally, the simulation engine uses efficient data structures so that you do not define millions of boxes by hand. Instead, you define the phantom as a single object, and the engine takes care of stepping from voxel to voxel as particles propagate.
In later chapters you will learn how to load image files, convert them into GATE voxel phantoms, assign materials and activity, and visualize the resulting geometry. For now, the key idea is that voxel phantoms translate image-based information into a physically meaningful 3D structure that radiation can interact with, which is the foundation of realistic, patient-oriented Monte Carlo simulations.
Views: 9