KAHIBARO
19.4. Scattered Coincidences
Table of Contents
Photon scattering before detection
Scattered coincidences in PET are coincidence events in which at least one of the two detected 511 keV photons has undergone one or more interactions, typically Compton scattering, before it is finally detected. Because of this scattering, the detected photon no longer travels along the original line of response that connected the annihilation point and the detectors. The event is still registered as a coincidence, but its geometric and energetic information is partially corrupted.
In GATE, scattered coincidences are not a different kind of particle. They are ordinary gamma photons whose interaction history includes one or more scatter processes in the patient, in surrounding objects, or even in the detector materials. The coincidence sorter will still pair the corresponding singles according to the timing window, but from the standpoint of image formation these events represent scattered coincidences and not true coincidences.
To understand how scattered coincidences appear in a simulation, it is useful to think in terms of the photon history. A positron emitter produces a positron that slows down and annihilates with an electron, emitting two photons with approximately 511 keV in almost opposite directions. As one of these photons crosses matter, it may undergo a Compton interaction described by electromagnetic physics. After Compton scattering the photon changes both its direction and energy. If the scattered photon still reaches the detector and deposits enough energy to satisfy the detector energy window, it can produce a valid single. When this single is paired with another single from the same annihilation, the resulting coincidence is classified as scattered in PET terminology.
In a GATE simulation, the physics list and production cuts control how accurately this scattering is modeled. Lower production cuts and appropriate electromagnetic models allow you to capture detailed scattering behavior inside complex phantoms and patient geometries. As a result, the singles and coincidences recorded by digitizers and actors will naturally include a mixture of true and scattered events that reflects realistic PET performance.
A scattered coincidence is a coincidence event where at least one detected 511 keV photon has scattered before detection, so the reconstructed line of response no longer passes through the true annihilation point.
From a data analysis perspective, scattered coincidences can be identified using the detailed information recorded in GATE output, especially in ROOT trees produced by hits, singles, and coincidence actors. Scattering flags or process information allow you to determine whether a detected photon experienced Compton scattering, how many scatter interactions occurred, and in which volumes they took place. This classification is central to PET performance studies, where you typically wish to compute scatter fractions and understand how phantom size, activity distribution, and scanner design influence scatter rates.
In practical PET simulations, scattered coincidences are reduced but not eliminated by applying an energy window around the 511 keV photopeak. Since Compton scattered photons lose energy, many of them fall below the lower bound of the window and are rejected at the singles level. However, some scattered photons will still fall within the window and contribute to scattered coincidences. When you design your GATE digitizer chain, the choice of energy thresholds therefore has a direct impact on the scattered coincidence fraction in the resulting data.
Accurately modeling scattered coincidences is essential whenever you want to evaluate correction algorithms, assess image quality, or compare PET system designs. A well configured GATE simulation that includes realistic geometries, materials, and physics will naturally produce scattered coincidences consistent with the physical processes of photon scattering before detection.
Views: 11