34.4. SPECT Simulation

Overview

Single Photon Emission Computed Tomography, usually called SPECT, is a nuclear medicine imaging technique that uses gamma emitting radionuclides and mechanical collimation to form 3D images of the activity distribution inside a patient. In Geant4, SPECT simulation means combining radioactive decay, gamma transport, a realistic collimator and detector model, and a way to record what a clinical camera would see. In this chapter the focus stays on the specific features and choices that are characteristic for SPECT, not on general medical physics or general Geant4 topics that other chapters already cover.

SPECT has some features that are different from many other Geant4 examples. The photon energies are usually in the range of about 70 keV to 300 keV. The detector is commonly a large planar or curved NaI(Tl) scintillation crystal viewed by many photomultiplier tubes. Spatial resolution is dominated by the collimator geometry rather than by the intrinsic detector resolution. In Geant4, these aspects guide how you design the geometry, choose physics models, define sources, and record data.

SPECT System Geometry in Geant4

At the core of a SPECT simulation is the camera head, which you typically place close to a volume that represents the patient or a test phantom. A basic Geant4 SPECT geometry usually contains, in order from the patient to the outside of the detector, a collimator, a scintillation crystal, a light guide or gap, and a mechanical housing. For many studies you can ignore optical photons and photomultiplier tubes and instead model only the gamma interactions in the crystal and the resulting idealized energy and position response.

The collimator is often the most geometrically complex part. Clinically used SPECT cameras typically use lead collimators with many parallel holes, but there are also pinhole and fan beam designs. To represent a parallel hole collimator, you can either build a detailed hole pattern or approximate it. A detailed model might use a repetitive structure, for example a hexagonal or square lattice of lead septa with air or vacuum holes. In Geant4 you can create this through parameterized volumes, replicas, or nested parameterizations, so that you do not have to define each hole by hand. You must ensure that the hole length, hole diameter, and septal thickness match the clinical or experimental design when your goal is quantitative accuracy.

Sometimes an analytic or semi analytic model of the collimator response is sufficient. In that case you may represent the collimator as a simple lead slab and apply angular and spatial acceptance criteria later in the analysis, or even after exporting simulated data. This can be useful for scans or sensitivity studies where a full geometric model would be too slow.

The scintillation crystal is usually a large NaI(Tl) block with a thickness of about 3/8 inch or 1/2 inch. In Geant4 this is just a box made of the appropriate scintillator material placed immediately behind the collimator. If you do not simulate optical photons, you typically treat any gamma energy deposited in this volume as measured energy, with later smearing to model resolution. For more sophisticated work, you can also include a backing material, such as aluminum or light guides, and sometimes a dead layer at the crystal surface to match measured efficiency and spatial response.

The patient or phantom can be represented in different ways. For simple studies, a water sphere or cylinder with embedded hot and cold regions is enough. For more realistic SPECT simulation, voxelized geometries derived from CT or anthropomorphic phantoms let you represent heterogeneous attenuation and scattering. In Geant4, such voxel phantoms are often built using parameterized volumes that assign different tissues and attenuation properties to each voxel.

Finally, you need to define a world volume that contains the patient, the SPECT camera head, and enough outer space so that scattered photons leaving the patient can still be tracked until they are absorbed or leave the scene.

SPECT Physics and Energy Range

SPECT uses single gamma photons, so you focus on electromagnetic interactions in the relevant energy range. Typical SPECT isotopes like $^{99\mathrm{m}}$Tc, $^{123}$I, or $^{201}$Tl emit photons with energies from about 70 keV up to 364 keV. At these energies, photoelectric absorption, Compton scattering, and coherent scattering in tissue and detector materials are the main processes.

In Geant4, you choose a physics list that provides accurate electromagnetic processes at low energies and that includes Rayleigh (coherent) scattering and atomic relaxation if you need fine spectral detail. For SPECT, common choices are the electromagnetic “option 3” or “option 4” models and low energy extensions. These models provide detailed cross sections for soft tissue, bone, lung, and NaI, and they reproduce the energy and angular distributions of scattered photons with good accuracy.

Production cuts determine when secondary particles like electrons are explicitly simulated. In SPECT, you generally want moderate cuts in tissue, since detailed electron tracks are not the primary interest, but you may use tighter cuts in the detector crystal to capture energy deposition accurately. It is important to ensure that gamma attenuation and scattering in both patient and collimator are modeled correctly. For radionuclide simulations where the decay scheme is important, such as simulations that start from the actual nuclear decay rather than a monoenergetic gamma beam, you also enable Geant4 radioactive decay physics and appropriate data libraries.

If you want to study detector effects such as scattered photons within the detector head or backscatter from the camera housing, you need to include all relevant materials with correct densities and compositions. The correct representation of lead collimators, NaI crystals, aluminum covers, and possibly photomultiplier glass affects the number and distribution of detected and scattered events.

Modeling SPECT Sources

A SPECT simulation can start from a description of the radionuclide distribution inside the patient or phantom. In Geant4 you have two basic approaches. You can explicitly model the nuclear decay using the radioactive decay process, or you can generate gamma photons directly with the General Particle Source or a custom generator.

Using explicit radioactive decay has the advantage that the full decay scheme is represented automatically. For example, $^{99\mathrm{m}}$Tc emits a 140 keV gamma and sometimes additional low energy photons or conversion electrons. When you use Geant4 radioactive decay, the code samples the appropriate emissions based on nuclear data. This is useful if you care about low energy photons or internal dose. It is also straightforward to define an activity distribution by assigning the radionuclide to different regions of the geometry.

On the other hand, if you are primarily interested in the imaging of the main photopeak, it is often simpler and faster to generate monoenergetic gamma photons with a specified energy that corresponds to the photopeak of interest, for example 140 keV or 159 keV. In that case you use the General Particle Source to define an energy spectrum, an emission position distribution inside the phantom, and an angular distribution. If the activity is uniformly distributed in an organ or phantom volume, you can configure a volume source and sample emission positions homogeneously. If you want to simulate a point or line source, for example in a test phantom, you set up a point or cylindrical source geometry.

The emission direction distribution can be isotropic, which corresponds to random decay directions in tissue. For some specialized studies you may restrict the emission directions, but this is less common in standard SPECT imaging simulations. When you simulate multiple energy windows or multi isotope imaging, you can define several sources or energy distributions and label events to keep track of their origin.

For time dependent simulations, such as dynamic SPECT acquisitions, you can couple the event generation to time, but for basic simulations you usually consider static distributions and integrate over time, since clinical scans last several minutes and the spatial activity distribution is often approximated as constant during that time.

Recording SPECT Detector Response

A key aspect of SPECT simulation in Geant4 is how you transform the underlying physics interactions into something that looks like what a clinical imaging system would measure. In practice, a SPECT camera records, for each detected gamma, an energy estimate, a position in the detector plane, and a projection angle or camera orientation. In Geant4 you extract these quantities from the energy deposited in the scintillation crystal and from the known detector geometry and acquisition parameters.

Usually you assign a sensitive detector to the scintillation crystal logical volume. Each Geant4 step in this volume can deposit some energy, and you accumulate these deposits for each event. For SPECT, an “event” in Geant4 can correspond to a single gamma emission. During that event, the gamma may undergo multiple interactions inside the crystal, and possibly create secondary electrons that deposit energy. You sum all energy deposits in the detector crystal for that event to obtain the total deposited energy, which you treat as the gamma energy measured by the camera, up to resolution effects.

To obtain a detector position for each detected photon, you can use different levels of approximation. The simplest approach is to take the position of the first interaction in the crystal as the “measured” position. Another common method is to use the energy weighted average of all interaction points in the crystal for that event, which approximates the light distribution centroid seen by the photomultipliers. For studies focused on system level performance, these simple approaches are usually sufficient.

You then apply detector energy resolution. SPECT crystals and electronics have an energy resolution that can be approximated by a Gaussian with a standard deviation proportional to the square root of energy or by a simple fractional resolution, for example 10 percent at 140 keV. In Geant4, after you have the true deposited energy $E$, you can sample a smeared energy $E'$ from a Gaussian distribution. For example, with fractional resolution $R$ (full width at half maximum divided by peak energy), the Gaussian standard deviation is often taken as $\sigma = R E / 2.355$. You implement this smearing in your analysis or in a user action class rather than in the core physics list. Similarly, you may add spatial smearing to the reconstructed position, for example by adding Gaussian noise with a width that represents the intrinsic detector resolution.

You also apply an energy window around the photopeak. Clinically, a SPECT camera would accept only events whose measured energy lies within a window, for example from 126 keV to 154 keV for a 140 keV photopeak. After you smear the simulated energies, you only record or use events that fall inside this window. Events outside the window represent scattered photons or noise and may either be discarded or used in scatter correction studies.

The detector orientation relative to the patient changes as the camera rotates around the patient during acquisition. In a Geant4 simulation, you can model this by rotating the detector head geometry around the phantom between runs or within a run by adjusting placement transforms. You record the current projection angle along with each detected event so that you can later bin events into sinograms or projection images.

From Geant4 Output to SPECT Images

Once you have simulated a set of detected events with energies, positions in the detector plane, and projection angles, you can create data products that resemble clinical SPECT projections. For each camera angle, you define a 2D grid of detector pixels, and you bin each accepted event into the appropriate pixel based on its position. The result is a set of 2D projection images at different angles that you can store as arrays or export to external tools.

In Geant4, you can collect this information using the analysis manager. You may define histograms for energy spectra to check your photopeak selection, as well as 2D histograms or ntuples with coordinates such as detector $x$, detector $y$, angle, and measured energy. For tomographic reconstruction, it is usually convenient to export an ntuple where each row contains at least the detector coordinates, angle, and energy. External software can then read this file, apply additional corrections, and perform image reconstruction using methods such as filtered backprojection or iterative reconstruction.

SPECT simulation is often used to study image quality, resolution, contrast, and quantification as a function of collimator design, detector properties, and clinical protocol. With Geant4 data, you can create projection datasets for different system configurations and then reconstruct images to compare spatial resolution, contrast recovery, and noise. Because Geant4 provides ground truth activity distributions, you can directly measure how well reconstructed images match the known distribution.

Another important use of SPECT simulation is to study attenuation and scatter correction. Since Geant4 tracks every photon and interaction, you know whether a detected photon was scattered in the patient or collimator and by how much. You can label events according to their scatter order and origin. In analysis, you can compare images built from only unscattered photons to images that include scatter, or you can design scatter correction algorithms and test them on the simulated data.

Finally, SPECT simulation can support system design studies. By changing collimator hole size, crystal thickness, or distance between patient and detector in the geometry, you can observe how sensitivity and resolution change. Geant4 provides a way to quantify these trade offs before building hardware. The same approach can be applied to novel collimator designs, multi pinhole systems, or hybrid SPECT and CT geometries.

By combining an appropriate source model, realistic electromagnetic physics, a carefully designed camera geometry, and a suitable way to record and process detector responses, Geant4 lets you build SPECT simulations that are useful for research, protocol optimization, and understanding the limitations and potential improvements of clinical imaging systems.

Views: 9