34.3. PET Simulation

PET Imaging within Geant4-based Medical Physics

Positron emission tomography, usually called PET, is one of the most common and important applications of Geant4 in medical physics. In this chapter, the goal is not to teach every technical implementation detail, because a dedicated practical PET example later in the course will do that. Instead, this chapter explains what is specific to PET simulations, how PET physics and detector behavior map to Geant4 concepts, and what particular choices you must think about when you design a PET study.

Clinical and Research Context of PET Simulation

In clinical practice PET is used to image metabolic and functional processes in the body. A radiotracer, often fluorodeoxyglucose (FDG) labeled with $^{18}$F, is injected into the patient. The tracer accumulates in tissues according to biological processes such as glucose uptake. PET then measures where annihilation photons are detected and reconstructs a three dimensional image of tracer distribution.

Geant4-based PET simulations are used to answer questions that are difficult, expensive, or impossible to explore experimentally. For example, a researcher may want to test a new detector geometry, compare different scintillator materials, or estimate the dose delivered by a particular PET protocol. PET simulations are also important for quantifying image quality, investigating artifacts, testing reconstruction algorithms, and designing time of flight PET systems.

Within the broader category of medical physics applications, PET simulation is unique because it combines nuclear decay, charged particle transport, gamma interactions, detector response, timing, and coincidence logic. A full PET simulation model often includes patient or phantom geometry, the PET scanner hardware, the radiotracer distribution, and data processing from individual interactions up to reconstructed images.

Representing Positron Emission and Annihilation

A central feature of PET is the decay of a positron emitting radionuclide and the subsequent annihilation of the emitted positron. Geant4 can represent this chain of processes with varying degrees of detail.

In a simplified PET simulation you may start the simulation from the annihilation photons only. That means you directly generate two 511 keV gamma rays with the appropriate angular correlation, usually back to back, and place them at the location of the radiotracer. This approach ignores the finite positron range and ignores non collinearity effects, but it is computationally efficient and can be useful for first studies of scanner geometry and idealized detector response.

In a more realistic PET simulation you let Geant4 handle the full radioactive decay and the transport of the emitted positron. You choose a positron emitting isotope, such as $^{18}$F or $^{11}$C, and use an appropriate physics list or decay module to model beta plus decay. The emitted positron travels through tissue, loses energy, and eventually annihilates with an electron, producing two photons. This approach naturally includes the positron range, that is the spatial distance between the emission point and the annihilation point, which can blur the image. It can also include non collinearity, which is a small deviation from perfect 180 degree emission due to residual momentum of the annihilating pair.

When you model full decay chains you must consider where and how you sample the decay positions. You can define the activity distribution using volumes that represent organs, lesions, or uniform phantoms. You can also incorporate time dependent activity or multiple tracers if needed. These choices affect the realism and complexity of the simulation.

PET Detector Modeling

PET detectors are typically made of scintillator crystals coupled to photodetectors such as photomultiplier tubes or silicon photomultipliers. In Geant4 you must decide how detailed your detector model should be.

At the geometry level, you represent individual crystals or blocks of crystals as volumes. Common scintillators in PET include LYSO, LSO, BGO, and sometimes others. Each material has characteristic properties such as density, effective atomic number, and attenuation coefficients. You can choose materials directly from the Geant4 NIST database or define them explicitly if needed for specific compositions.

At the physics level, PET relies mainly on electromagnetic interactions of 511 keV photons. The prominent processes are the photoelectric effect and Compton scattering. A detected event corresponds to a gamma photon interacting in a crystal and depositing sufficient energy to generate a measurable light signal. In your simulation you often treat the deposited energy in the scintillator as a proxy for the detector output and apply additional modeling for energy and time resolution.

If you want to study high granularity scanners, depth of interaction capabilities, or parallax effects, you need a fine segmentation of the detector geometry. This usually means arrays of small crystals arranged in rings or axial blocks around the imaging field of view. You will often assign unique identifiers to each crystal, such as ring and crystal indices, so you can reconstruct which detector element registered a hit.

Coincidence Detection Concept

PET does not use the absolute rate of detected photons directly. Instead, it mostly uses coincident detections. A true PET event is a pair of annihilation photons detected nearly simultaneously in opposite detector elements. The line joining the two detection positions is called a line of response and is used for image reconstruction.

In simulation, coincidence logic is applied to detector hits or to processed detector events. You define a coincidence time window, for example a few nanoseconds, and consider two detector signals as a coincidence if their detection times differ by less than this window and if they originate from the same annihilation interaction. Geant4 itself transports particles and records hits, but the grouping of hits into coincidences is performed in your user code or in a later analysis step.

The types of coincidences that matter in PET are called true, scatter, and random coincidences. True coincidences come from the two original annihilation photons detected without significant scattering that changes their direction. Scatter coincidences involve at least one photon that has scattered in the patient or in the detector before detection, which can misplace the line of response. Random coincidences occur when unrelated photons from different annihilations are detected within the coincidence window, which adds uniform background to the reconstructed image.

Geant4 allows you to identify these categories if you keep track of particle history and processes. For example you can label coincident pairs as true if both photons have a direct ancestry from the same positron annihilation and minimal scattering, or as scattered if one has undergone Compton scattering in the phantom, or as random if the pair originates from different primary decays. This classification is important for evaluating scanner performance and for testing correction techniques.

Time of Flight PET Considerations

Modern PET scanners often use time of flight information. Time of flight PET measures the slight difference in arrival times of the two annihilation photons. From this, it estimates where along the line of response the annihilation occurred. This can significantly improve image signal to noise ratio, especially for large patients.

Within Geant4, time of flight PET can be studied only if you pay attention to timing information in the simulation. Geant4 assigns a global time to each step and to each track, and you can derive detector signal times from the arrival times of photons and from models of scintillation light production and photodetector response. In many PET simulations a simplified timing model is used, where the detection time is taken as the time when most of the energy is deposited or when the first interaction occurs, plus a stochastic timing jitter that represents the detector time resolution.

To investigate time of flight performance you simulate many annihilation events, extract the times at which each detector element registers a signal, and study the distribution of time differences between coincident hits. The width of this distribution reflects both the intrinsic physics, such as path length differences, and the detector timing characteristics that you impose in your model.

Dose and Radiation Safety Aspects of PET

Although PET radiotracers are typically used in diagnostic, not therapeutic, contexts, dose estimation remains important. Geant4 can provide detailed absorbed dose distributions in patient or phantom models for a given PET protocol. For example, you might want to evaluate the dose to specific organs from repeated scans or from new tracers with different biokinetics.

In a PET simulation you can assign biological tissue materials to organ geometries and accumulate energy deposition from all particles originating from the radionuclide decays. This includes contributions from positrons, annihilation photons, and any accompanying gammas or X rays. By normalizing the simulated dose to injected activity and acquisition time, you can derive estimates of organ and effective doses.

From a radiation protection perspective, Geant4 PET simulations can also model radiation fields around the scanner, dose to staff, or shielding requirements in a PET facility. This is done by extending the simulated geometry to include the scanner room, walls, and shielding, and by tallying doses at relevant locations.

Image Quality Metrics from Simulated PET Data

One of the main motivations for PET simulation is to study image quality. Given that Geant4 produces event level information, you can feed this simulated data into reconstruction software and then analyze the reconstructed images. Common image quality metrics in PET include contrast recovery, signal to noise ratio, spatial resolution, and quantitative accuracy of standardized uptake values.

To bridge from Geant4 output to image quality you first process hits into coincidences and lines of response. Next you bin them into sinograms or list mode data that your reconstruction software expects. After reconstruction you compare known phantom activity distributions with the reconstructed images. Because you control the ground truth in simulation, you can quantify how different scanner designs, acquisition protocols, or correction methods influence the image.

For example you can simulate a standard PET image quality phantom with hot and cold spheres and measure how well different scanner geometries recover the contrast of small lesions. You can also vary reconstruction parameters or the number of events to examine the trade off between scan time and image noise.

Specific Challenges and Simplifications in PET Modeling

PET simulations can easily become complex and computationally intensive. As a result, you often have to choose appropriate simplifications. A few common decisions are whether to include optical photon transport, whether to simulate detailed electronics, and how realistic to make the patient or phantom model.

Full optical simulation, with scintillation photon generation, transport, reflections, and photodetector response, can yield very accurate predictions of energy and timing resolution but is costly. Many PET studies instead use a simplified detector model, where energy deposition in crystals is converted directly into measured energy and timing by applying analytic smearing functions for energy and time resolution.

Electronics modeling can be simplified by assuming perfect dead time and pile up behavior or by ignoring these effects for low count rate studies. For high activity protocols or time of flight scanners, more realistic electronics models may be necessary.

Patient anatomy can be represented by simple uniform cylinders or by realistic voxelized phantoms derived from CT data. Simple geometries are suitable for method development, while detailed anthropomorphic models are more appropriate for dosimetry or protocol optimization.

The art of PET simulation is to choose a level of detail that matches your research question. Geant4 provides the physics and geometry capabilities, but you must decide which parts of the system are modeled precisely and which are approximated.

Role of PET within the Wider Geant4 Medical Physics Toolkit

Within the landscape of Geant4 medical physics applications, PET sits alongside SPECT, radiotherapy, and proton therapy. PET simulations emphasize diagnostic imaging, coincidence detection, and tracer kinetics, while radiotherapy simulations focus on dose and biological effect, and proton therapy simulations emphasize charged particle transport and Bragg peak behavior.

Skills and tools developed in PET simulations, such as handling radioactive decays, modeling detector arrays, and processing large volumes of event data, transfer to many other applications. In particular, the way PET uses coincidences, time of flight, and detailed detector modeling sets a pattern that can inspire designs for Compton cameras, prompt gamma imaging, and other advanced modalities.

As you progress through this course, the later practical chapters will guide you step by step through building an example PET scanner in Geant4. The conceptual understanding from this chapter will help you make sense of the choices you encounter in that implementation and adapt them to your own research problems.

Views: 11