KAHIBARO
27.3. Creating Detector Crystals
Table of Contents
Crystal size
Detector crystals are the smallest repeating sensitive elements in a PET scanner. In GATE, each crystal is a geometry volume with a specific size, position, material, and identification. The ring and block structures are handled in other sections, so here the focus is only on defining individual crystal dimensions and how that choice affects the simulation.
For a PET crystal, you define three linear dimensions: length (axial), width (transaxial, along the ring circumference), and thickness (radial, from the patient toward the outside of the scanner). In OpenGATE Python, these map to the size settings of a box volume, for example by assigning a list like [size_x, size_y, size_z] with each value combined with the correct unit, such as 10 mm or 2 cm. The absolute values you choose must be consistent with the ring diameter and block layout that are defined elsewhere, so that crystals fit properly inside their modules and the ring closes without overlaps.
Typical PET crystals are long in the axial direction, moderately thick radially, and narrow in width. A conventional whole body PET scanner might use crystal cross sections of about $4 \times 4$ mm and lengths of 20 mm or more. Preclinical systems for small animals usually use smaller cross sections and shorter lengths to achieve higher spatial resolution. If your array has $N$ crystals along a particular dimension, the total block dimension in that direction will be approximately $N$ times the crystal size plus any gaps. This is important when you later repeat crystals to form modules or rings, since you must choose repetition pitches that match the crystal size plus the intended gap, not simply the bare crystal size.
Choosing crystal size affects both physics and computing cost. Larger crystals collect more signal per element but reduce spatial resolution. Smaller crystals increase the number of channels and therefore the number of volumes in the simulation, which may slow down geometry navigation and output handling. For a first PET model, it is often practical to use a slightly simplified crystal grid with fewer, slightly larger crystals, so that simulations run faster while you learn. Later you can refine the dimensions to match a real scanner.
In many PET geometries, crystals are defined as box solids where one axis is aligned radially. That axis, sometimes called the thickness, controls how much scintillator material photons traverse before escaping. Thicker crystals improve detection efficiency at 511 keV but may increase parallax effects and degrade radial resolution. You set this explicitly in the crystal volume size. The other two axes define the footprint of each crystal at the inner surface of the ring and the axial coverage. Matching these to the detector ring parameters ensures that crystals line up correctly around the ring and along the axial extent.
When you construct the geometry in GATE, you do not need to explicitly encode physical concepts like width or thickness in the code, but it helps to maintain a clear convention and document it. For example, you might decide that size_x is crystal width, size_y is crystal thickness, and size_z is axial length, and then consistently use that convention for all PET components. This makes later steps such as assigning detector IDs or projecting lines of response easier to reason about.
Crystals should have no overlaps with neighboring crystals or with the world boundary. When arranging many crystals, small numerical errors can accumulate, especially if you mix units like mm and cm. Always use the units from the GATE units chapter consistently when defining sizes. After you define crystal size, you can rely on GATE geometry visualization to verify that crystals are correctly sized and that the array looks as expected.
Keep crystal sizes consistent with ring and block dimensions, use a clear axis convention, and avoid overlaps by matching repetition pitch to crystal size plus any intentional gap.
Crystal material
Each PET crystal must be assigned a scintillator material. The choice of material determines key detector properties, such as stopping power at 511 keV, light yield, decay time, and energy resolution. In GATE, you represent these by selecting an appropriate predefined material or by defining a custom one, then assigning it as the material of the crystal volume.
For most modern clinical PET scanners, lutetium based crystals are used. A widely used example is LYSO, chemically Lu$_2$SiO$_5$:Ce. GATE typically provides standard detector materials like G4_LYSO or similarly named entries through the underlying Geant4 material database. Other materials that appear in PET literature include BGO, LSO, and GSO. Each material has different density and effective atomic number, which directly affect detector efficiency and the probability that 511 keV photons interact in the crystal rather than passing through.
When building a basic PET model, it is often sufficient to choose one of the common built in materials. You simply specify the name of the material in the crystal volume definition. For example, for LYSO, you might assign the material string that corresponds to the LYSO definition available in your GATE setup. The exact naming and availability follow the materials chapter and the Geant4 material database, so you should check those resources to confirm you are using the correct identifier.
In more advanced simulations, you might need crystal materials that are not provided by default, or you might want to refine properties such as density or elemental composition. In that case you define a custom material once, for example with a given composition in terms of elements and a specific density in g/cm$^3$, and then use that material for every crystal. The material definition itself belongs to the materials chapter, but from the crystal perspective the important point is that all crystals of the same type share the same material, so that the detector behaves uniformly.
The crystal material interacts strongly with the physics list. At 511 keV, higher density and higher atomic number materials provide higher detection probability but can also influence scatter fractions and energy deposition patterns. When combined with digitizers, the material indirectly affects observables like singles energy spectra and the shape and height of the 511 keV photopeak. If you later introduce energy blurring to simulate finite energy resolution, the underlying material choice still matters because it controls the true deposited energy distribution before blurring.
In a PET ring, you usually assign one primary scintillator material to all crystals. If the scanner includes specialized components like depth of interaction layers or different materials in inner and outer crystal rows, you can represent that by defining separate volume types with different materials and sizes, then arranging them appropriately in the geometry. Regardless of complexity, the rule remains that every crystal volume must have an explicitly defined material that matches the detector design you want to simulate.
Assign a physically realistic scintillator material, such as LYSO or BGO, to every crystal volume, and keep material definitions consistent across all crystals of the same type.
Views: 9