KAHIBARO
28.5. True Events
Table of Contents
Identifying true coincidences
In PET, a true coincidence is a pair of detected photons that both originate from the same positron annihilation and both reach the detectors without undergoing any interaction that changes their direction or energy significantly. True events are the useful signal in PET, because they correspond to a correct line of response between the two detection points.
In a GATE simulation you can exploit information that is not available in experiments, in particular the link between each detected photon and its original source particle. This allows you to label coincidences as true, scattered, or random on an event-by-event basis.
A PET coincidence in GATE is usually formed from two singles created by the digitizer chain. Each single corresponds to energy deposited in one detector crystal and already includes effects such as energy summation and blurring. When the coincidence sorter finds two singles inside the coincidence time window, it writes a coincidence entry that contains information from both singles, typically including identifiers such as event ID, source ID, and scattered flags. These identifiers are what you use to decide whether the coincidence is true.
To classify coincidences, you generally rely on three logical conditions. First, both photons must come from the same annihilation. In practice this is checked using a shared event identifier or parent ID that traces back to the positron or its annihilation. Second, there must be no nontrivial Compton or Rayleigh scattering in the object or detector that significantly changes their paths. In GATE this is often encoded by a flag indicating whether the gamma has scattered before detection. Third, both photons must pass the PET energy and timing selection you have applied, so that you are analyzing only coincidences within your chosen energy window and coincidence timing window.
The combination of these conditions lets you create a clean subset of coincidences that represent ideal PET signal. This is very useful when you evaluate scanner performance, for example when you calculate sensitivity or true fraction as a function of activity or when you compare different detector designs.
In ROOT or Python analysis you typically read the coincidence tree and build logical expressions that mark each event as true or not. You can store this as an extra boolean or integer field so that you can easily filter or histogram only true events. Since GATE gives you direct access to ground truth information, do not try to infer trues indirectly from energy alone; explicitly using the source and scatter identifiers is more reliable and easier to reproduce.
A coincidence is usually counted as a true event if and only if:
- The two photons originate from the same positron annihilation (same event or parent ID).
- Neither photon has scattered before detection (no Compton or Rayleigh interaction on its history flag).
- Both photons satisfy the selected energy window and coincidence timing window of the PET system.
These conditions define the ideal PET signal used in performance studies and algorithm validation.
Views: 9