KAHIBARO
36.2. Voxelized Activity Maps
Table of Contents
Patient-specific activity
In internal dosimetry you often want to compute dose for a specific patient instead of using simple idealized phantoms. Patient-specific activity describes how a radiopharmaceutical is distributed inside that one patient as a function of space and time. GATE uses voxelized activity maps to represent this kind of information.
A voxelized activity map is a 3D grid where each small cube, or voxel, has an assigned activity value, typically in units of Bq or Bq per voxel. In practice, these maps usually come from nuclear medicine images such as SPECT or PET, sometimes combined with anatomical images such as CT or MRI for geometry and materials.
Inside GATE, the key idea is that the source is no longer a simple point or volume with uniform emission. Instead, the probability that a radioactive decay occurs in a particular voxel is proportional to the activity stored in that voxel. During the simulation, events are sampled according to this spatial distribution. If a time dependence is included, for example several activity maps at different time points, GATE can also model how the source distribution evolves.
It is important to distinguish between activity and activity concentration. Activity describes the total number of decays per second in a voxel. Activity concentration describes activity per unit volume or per unit mass, for example Bq per milliliter or Bq per gram. When you build a voxelized source you must know which one your images represent, since this will determine how you convert voxel values into absolute activity. GATE expects a clear convention, so any scaling between image values and physical activity must be applied consistently before or during source definition.
Another important consideration is alignment. The voxelized activity map must be placed inside the simulation geometry so that it corresponds to the actual patient anatomy. The image origin, spacing, and orientation must be coherent with the patient geometry that you use for dose scoring. Typically, the same grid can be used both for activity and for dose, but they can also be separate. In either case, misalignment between the activity map and the patient materials will lead to incorrect dose estimation.
In patient-specific dosimetry, activity is also a function of time. A single static map corresponds to one time point, often at or near the time of imaging. For accurate dosimetry you normally require the time-integrated activity, which describes the total number of decays that occur in each voxel over the whole therapy period. You can approximate this in GATE by using multiple simulations with different activity distributions and durations, or by scaling a single simulation using precomputed time-integrated activity coefficients.
The choice of temporal model is critical. If you assume a constant activity in time, you may underestimate or overestimate dose in organs where uptake or clearance is fast. On the other hand, using several time points increases complexity and simulation effort. In practice, the approach often depends on the available imaging data and the clinical question. For method development, you may start with a single static activity map to validate geometry and physics, then move to more realistic time-dependent activity once the basic setup is correct.
For patient-specific internal dosimetry, the dose in a voxel is proportional to the time-integrated activity in that voxel, not only to the activity at a single imaging time point.
When you construct a voxelized activity map in GATE, always verify three things. First, the total activity represented by the map must match the intended administered or organ-specific activity after any decay correction. Second, the spatial distribution must be visually consistent with the underlying anatomy. Third, the resolution, that is voxel size, must be appropriate for the organ structures and for the expected dose gradients. Very large voxels can smooth out important details, while very small voxels may greatly increase computing time without significantly improving accuracy.
Source images
Voxelized activity maps in GATE are usually derived from source images. These images come from clinical or preclinical SPECT or PET scanners, or from synthetic distributions produced by computational phantoms. In all cases, the image grid defines where sources are located and the voxel values define how strong the source is in each position.
Source images are commonly stored in formats such as DICOM, NIfTI, or MHD. The chosen format controls how you access image metadata needed for the simulation. You must know at least the image dimensions in voxels, the voxel spacing in millimeters, and the image origin in patient coordinates. These three pieces of information determine the physical size and placement of the activity map in the simulated world.
When using clinical nuclear medicine images as source images, raw reconstructed images are usually in units related to counts or counts per second, not directly in Bq. To use them as activity maps you must convert them. This conversion can involve scanner calibration factors, acquisition time, and decay correction applied during reconstruction. After conversion, voxel values should represent physical activity or activity concentration at a known reference time.
If your source images represent activity concentration, for example Bq per milliliter, you must multiply each voxel value by the corresponding voxel volume to obtain activity in Bq per voxel. The voxel volume is given by the product of the three voxel spacings in millimeters, converted to milliliters. If your source images already contain activity in Bq per voxel, you can use them directly, but you still have to check whether any global scale factor was applied during reconstruction or pre-processing.
In many workflows, different images are used for different tasks. A CT image defines materials and density for the patient geometry, while a SPECT or PET image defines the activity distribution. In GATE, these can be independent image-based actors and volumes, as long as their coordinate systems are consistent. To prevent misalignment, it is good practice to read both images in the same coordinate convention, verify their shared origin and orientation, and visualize them together in the simulation to ensure that activity appears in plausible anatomical locations such as organs or lesions.
Noise and reconstruction artifacts in source images can affect dose calculations. For instance, hot or cold spots caused by reconstruction noise can create unrealistic dose heterogeneity. Some workflows apply smoothing or segmentation to the source images before using them in GATE. For example, activity may be assumed uniform inside segmented organs and zero outside, even if the original image was noisy. This simplifies interpretation at the cost of losing fine spatial detail.
If multiple time points are available, such as dynamic PET or repeated SPECT scans, you may have a series of source images that each describe the activity distribution at a different time. In GATE, you can use such a series to model time-dependent activity by changing the map during the simulation or by running separate simulations for different time intervals and then summing the resulting doses. The choice depends on how you want to manage computational effort and on how the time axis is represented in your overall dosimetry model.
For synthetic studies, you can generate source images algorithmically. For example, a digital phantom can provide a labeled organ map, and you assign specific activity levels to each organ label to produce a 3D activity distribution. This type of source image is useful when you want full control over the ground truth activity and dose, for validation or method development. GATE treats these synthetic images in the same way as clinical ones, as long as the format and metadata are correctly specified.
When using source images in GATE, always ensure that:
- Voxel values are correctly converted to physical activity or activity concentration.
- Image spacing, origin, and orientation are consistent with the patient geometry.
- The total activity represented by the image matches the intended physical scenario.
Once the source images are correctly prepared, you can use them to define voxelized sources in GATE. The simulation then samples decay positions according to the image intensities, and all subsequent physics, tracking, and dose scoring proceed exactly as for any other source. This approach provides a direct bridge between clinical imaging data and patient-specific internal dose calculations.
Views: 11