KAHIBARO
42.3. Physics Validation
Table of Contents
Interaction processes
Physics validation checks whether the interactions and transport in your GATE simulation match what is expected from real physics and from the chosen physics list. For an absolute beginner, it is enough to understand that you must confirm that particles in your simulation interact in the right way, at the right frequency, and with the right angular and energy distributions.
A basic first step is to verify that the relevant processes are actually enabled. For example, in a PET or SPECT simulation you expect photoelectric effect, Compton scattering, and possibly pair production for higher energies. In proton therapy you expect ionization, multiple scattering, nuclear interactions, and production of secondary particles such as neutrons. You should check the physics configuration you selected and make sure that all processes that matter for your application are included, and that unnecessary ones are not accidentally disabled.
In practice, you can validate interaction processes by looking at the information stored in hits, phase space, or dedicated actors. Inspect the recorded interaction type or process name, the volume where the interaction occurred, and the particle type. For example, in a PET detector crystal, most 511 keV photons that contribute to useful events should undergo photoelectric absorption or a combination of Compton scatter followed by photoelectric absorption inside the crystal. If your output shows that almost all interactions are Compton and very few photoelectric events occur, this might indicate a problem with material definition, density, or the selected electromagnetic physics models.
Similarly, for attenuation studies in a shielding slab, you can compare the fraction of photons that pass through without interaction to analytical expectations based on the linear attenuation coefficient. The number and type of secondaries produced, such as scattered photons and electrons, should be reasonable relative to the known cross sections. In proton therapy, you can check the longitudinal profile of proton energy loss and the frequency of nuclear interactions against reference data or published results.
It is also important to confirm that spatial distributions of interactions are plausible. Interactions should be more frequent in high density or high atomic number materials compared to air or soft tissue. Within a voxelized patient geometry, interactions should cluster in bones and dense tissues for keV photons, or along the proton track for therapy beams. If you observe interactions in empty space, or the absence of interactions in regions with significant material, this points to possible geometry or material assignment errors rather than pure physics list issues, but it still belongs to physics validation because it affects how physics is realized in the model.
For many applications, angular distributions of scattered particles are relevant. For instance, the angular distribution of Compton scattered photons in a water phantom can be compared qualitatively with theoretical expectations from the Klein–Nishina formula. In proton therapy, the lateral spread due to multiple scattering can be compared with reference measurements or published Monte Carlo benchmarks.
Whenever possible, you should identify a simple geometry and source configuration where analytical or semi analytical results are available. For example, a monoenergetic gamma beam through a slab, or a proton pencil beam in water. Use these simplified setups to validate that the interaction processes and their spatial distributions are correct before moving to full scanner or patient simulations.
Important rule: Always validate interaction processes first in simple test geometries, using known or reference results, before trusting them in complex clinical or imaging simulations.
Energy spectra
Energy spectra are one of the most powerful tools for physics validation in GATE, because many physical effects show up very clearly in the distribution of detected or deposited energy. A spectrum is typically a histogram of counts versus energy, such as deposited energy in a detector crystal or energy of photons exiting a phantom.
The first task is to check that the main peaks appear at the correct energies. For a PET simulation using F-18, you expect a photopeak near 511 keV in the singles spectrum. For a Tc-99m gamma camera, the primary photopeak should be near 140 keV. In a megavoltage photon beam, you expect a broad spectrum with a characteristic maximum energy defined by the beam configuration. If the peak position is shifted, you might have used incorrect units, an incorrect source energy definition, or misconfigured energy blurring in the digitizer.
You should also examine the relative shape of the spectrum around the main peaks. For gamma detectors, you expect a photopeak, a Compton continuum, and possibly escape peaks or backscatter peaks depending on geometry and materials. Comparing these features to measured spectra from similar experimental setups or to published Monte Carlo spectra is a strong test of your physics configuration and material definitions. If the Compton continuum is too low or too high relative to the photopeak, you might need to review your electromagnetic physics list, production cuts, or geometry.
For attenuation and shielding simulations, transmitted spectra can be compared with the exponential attenuation law and with known buildup behavior. A slab of lead will reduce the intensity of the primary peak and enhance lower energy scattered photons emerging from the shield. If your output only shows a simple scaling of the primary peak with no scattered component when you expect one, you might be missing important scattering processes or have incorrect material composition.
In proton and electron beam simulations, the energy spectra of particles at different depths in a phantom provide insight into whether energy loss processes are modeled correctly. For protons in water, the mean energy should decrease with depth, with a characteristic distribution near the Bragg peak. Comparing these depth dependent spectra with reference data or benchmark codes is an important validation step.
It is also essential to validate how energy spectra change when you modify simulation parameters. For instance, if you change the detector energy resolution, the simulated photopeak should broaden in a way consistent with the specified Gaussian width. If you narrow the energy window in a PET or SPECT digitizer, the number of events in the accepted spectrum should change as expected, but the underlying physical spectrum produced before digitization should stay the same. This distinction helps separate physics validation from digitizer validation.
To connect spectra with analytical expectations, you can sometimes compute theoretical distributions. For monoenergetic gamma photons in a thin scatterer, you can estimate the expected Compton scattered energy distribution using the Compton formula
$$
E' = \frac{E}{1 + \frac{E}{m_e c^2}(1 - \cos\theta)}
$$
and compare qualitative features with your simulated scattered spectrum. You do not need exact agreement point by point, but overall trends such as the energy range and shape should be consistent.
Finally, when validating spectra, pay attention to statistical uncertainty. Noisy spectra with few events are hard to interpret. You should run enough histories so that the main peaks and continuum are smooth and stable, and then compare them to reference data.
Important rule: Validate that simulated energy spectra show peaks at correct energies, realistic shapes, and consistent changes when you modify physics or detector parameters. Incorrect spectra usually indicate problems in source definition, materials, or physics configuration.
Views: 10