KAHIBARO
28.6. Scatter Events
Table of Contents
Identifying scattered coincidences
In PET, a scattered coincidence is a coincidence event in which at least one of the detected photons has undergone one or more interactions that change its direction or energy before reaching the detector. These interactions typically occur through Compton scattering in the patient, in the scanner hardware, or in surrounding materials. Scattered coincidences usually still fall within the PET energy and timing windows, so they are recorded as valid events, but their lines of response no longer pass exactly through the true annihilation position. This degrades image contrast and quantitative accuracy.
In GATE, and in particular when you use the PET digitizer chain, scattered coincidences are not just defined by detector signals but can also be classified using Monte Carlo truth information. The most useful information for this is stored in the coincidence output, which normally includes flags or labels that indicate whether each coincidence is true, scattered, or random. Internally, these labels are decided by tracing the history of the two photons that form the coincidence back to the original annihilation and checking whether any non trivial scattering interaction occurred along their paths.
The basic PET logic is that a coincidence is marked as scattered if at least one of the two photons is scattered. A single Compton scatter in soft tissue is enough to change the emission direction and energy. Even if the final deposited energy in the detector crystal is still inside the PET photopeak window, the geometrical relationship between the two detected crystals and the annihilation point is altered. When you reconstruct images from such coincidences, activity is redistributed away from its origin and appears as scattered background.
From the point of view of Monte Carlo truth, GATE records detailed information about each photon, such as its interaction processes and step history. The coincidence module can use this history to set a scatter classification. For example, if the photon took a straight path from the annihilation to the detector without any scattering interaction, it belongs to a true coincidence. If, along its track, the photon had a Compton interaction in the patient phantom, then continued with a new direction and finally interacted in a detector crystal, the associated coincidence is classified as scattered. If both photons are scattered, the event is still a scattered coincidence, with even poorer localization.
In practice, when you read GATE coincidence output, you will often see fields such as a scatter flag, a scatter type, or a scattered volume identifier. These allow you to select only non scattered events for ideal reconstructions, or to quantify how much scattered background you have under your energy window. For beginners, the main idea is that you rarely need to reconstruct from all coincidences. Instead, you filter the coincidence list based on those scatter labels to remove as many scattered events as possible, or to study how scatter fraction changes when you modify the scanner geometry, energy window, or phantom.
It is important to distinguish scattered coincidences from random coincidences. Both reduce image quality, but random coincidences come from two independent annihilation events that happen to fall inside the coincidence timing window, while scattered coincidences still originate from a single annihilation but are geometrically distorted by scattering along the photon paths. GATE can classify both types separately through its coincidence sorting and event history, allowing you to study them independently.
In PET performance studies, you will often compute a scatter fraction, defined as the fraction of all detected true plus scattered coincidences that are due to scatter. GATE is particularly useful here because it can give you an accurate count of true, scattered, and random events from the Monte Carlo ground truth, which you can compare with experimental estimates. When you later move on to reconstruction or to system optimization, this ability to identify and extract scattered coincidences from the simulated data is essential for understanding system performance and for validating scatter correction methods.
A PET coincidence is classified as a scattered coincidence if at least one of the two detected photons has undergone one or more scattering interactions before detection. In GATE, this classification is based on the full Monte Carlo history of each photon, not just on its final energy or position.
Views: 10