KAHIBARO
23.3. Material Mapping
Table of Contents
Hounsfield units
In voxelized geometry, each voxel in a CT image carries a number called a Hounsfield unit, often abbreviated as HU. This number represents how strongly that voxel attenuates x rays, relative to water. HU are derived from the linear attenuation coefficient $\mu$ that a CT scanner reconstructs from the projection data.
The HU scale is defined by water and air as reference points. Water is set to 0 HU, air is set to −1000 HU, and other tissues fall somewhere in between or above. The definition is
$$
\text{HU} = 1000 \times \frac{\mu_\text{voxel} - \mu_\text{water}}{\mu_\text{water}}
$$
Here $\mu_\text{voxel}$ is the linear attenuation coefficient of the material represented by the voxel, and $\mu_\text{water}$ is that of water, both at the x ray energy spectrum of the CT scanner.
Typical HU values for some simple materials are:
| Material | Approximate HU |
|---|---|
| Air | −1000 |
| Lung (low density) | −800 to −500 |
| Fat | −150 to −50 |
| Water | 0 |
| Soft tissue | 20 to 80 |
| Cancellous bone | 100 to 300 |
| Cortical bone | 700 to 1500 |
CT systems output HU as integer values stored in each voxel of the image. When you use a CT image to drive a GATE simulation, these HU values must be converted to materials and densities that Geant4 can understand. HU are not used directly by the physics engine. Instead, an intermediate conversion step produces a material label and density for each voxel.
In practice, this conversion depends on how the CT scanner was calibrated. A standard way to obtain a mapping is to scan a calibration phantom, which contains known materials at known densities. By measuring their HU values and comparing them to reference attenuation or density values, you can construct calibration curves. These curves relate HU to physical quantities such as electron density or mass density.
In GATE, the HU information is read from the CT image through a DICOM or image reader. The voxel HU values are then passed to a CT to material module that applies such calibration relationships. This tool assigns each voxel both a material identity and, in many cases, a density. Later, the physics models use these assignments to compute interaction probabilities and energy deposition.
It is important to remember that HU depend on the x ray spectrum used during CT acquisition. A given tissue may have slightly different HU values on scanners with different tube potentials or reconstruction algorithms. For simple training simulations, you may use generic HU to material tables. For accurate patient dosimetry and quantitative imaging studies, you should rely on scanner specific calibration data.
Tissue materials
To run a voxel based simulation, GATE must assign each voxel a Geant4 material. The material determines the elemental composition and mass density that the physics models use. Material mapping converts HU into such tissue materials. This mapping can be as simple or as detailed as your application requires.
The basic idea is to define a set of tissue classes, such as air, lung, fat, soft tissue, and bone, and to assign each HU range to one of these classes. For example, all voxels with HU less than −900 can be labeled as air. Voxels between −900 and −300 can be labeled as lung tissue. HU close to zero may represent water or generic soft tissue. High HU values may represent different kinds of bone.
A simple HU to material table might look like this:
| HU range | Assigned material | Typical density (g/cm³) |
|---|---|---|
| HU ≤ −950 | Air | 0.0012 |
| −950 < HU ≤ −300 | Lung | 0.2 to 0.4 |
| −300 < HU ≤ −50 | Adipose (fat) | 0.9 |
| −50 < HU ≤ 150 | Soft tissue / water | 1.0 |
| 150 < HU ≤ 700 | Cancellous bone | 1.2 |
| HU > 700 | Cortical bone | 1.8 to 2.0 |
In practice, you can use more categories if you want to distinguish, for example, muscle, brain, liver, cartilage, and different bone types. Each category is associated with a Geant4 material that specifies its elemental composition, for example mass fractions of H, C, N, O, Ca, and P, and an assigned density.
In GATE, material mapping often separates two tasks. First, HU are converted to a mass density, usually with a continuous function. Second, that density is grouped into a material class, or the same base composition is scaled by density.
A simple density mapping can be linear over certain HU ranges. For instance, for soft tissues, you might apply
$$
\rho(\text{HU}) = \rho_\text{water} \left(1 + \alpha \frac{\text{HU}}{1000}\right)
$$
where $\rho_\text{water}$ is 1.0 g/cm³ and $\alpha$ is a calibration factor derived from a phantom scan. For bone, you might use a different relation that better matches the higher HU range and the way bone density increases with HU.
In GATE, you can implement this mapping using a CT to material configuration. Typically, you provide a file or a Python structure that links HU or density ranges to material names. The tool then reads the CT image voxel by voxel, calculates the density from HU, then assigns the voxel one of the predefined materials according to the specified rules.
When building such a map, you must think about the level of detail you need. For a simple scatter study, grouping all non bone voxels into a single soft tissue material might be sufficient. For internal dosimetry, detailed organ specific compositions may be necessary, for example, specific materials for liver, kidneys, and bone marrow. These materials can be taken from standard databases, such as ICRU or ICRP reference tissues, and registered in GATE as custom materials.
Consistency between the CT calibration, the HU to density relationship, and the material table is essential. If you use a density relation derived from one scanner and a material table intended for another, your simulated tissues may not match real patient properties. For clinical or research studies that rely on accurate dose or quantitative imaging, this can lead to errors.
Finally, after material mapping, it is important to visualize and inspect the voxelized geometry. In GATE visualization, you can color each material differently to check whether air, lung, soft tissue, and bone appear in the correct anatomical regions. This quality check helps catch mistakes in HU thresholds, sign conventions, or CT scaling before running long simulations.
Views: 11