KAHIBARO
27.6. Configuring the Digitizer
Table of Contents
Singles
In a PET scanner simulation, the digitizer is the part of GATE that transforms raw hits in the detector crystals into signals that look like what the electronics would output. The first key product of this chain is the single event, usually called a “single.” A single corresponds to one detection in one detector channel after basic processing steps such as energy summation and time assignment.
In the PET project, you first ensure that hits are produced in the crystals by a hits actor. The digitizer then groups the energy deposits that belong to the same physical detection. The typical configuration is to create a singles module in the digitizer chain and attach it to the volume where hits are scored, usually the crystals. In OpenGATE this is done by creating a digitizer object, then adding a singles stage that reads the hit collection. You specify which attributes should be kept in the singles, for example energy, position, time, and detector identifiers.
The singles stage defines what one detector event means. Usually, it sums all energy deposits that occur in the same detector element within a short time coincidence window at the electronics level. The result is one single event with a single energy and time stamp for that detector element. For PET you also want to carry information about the detector structure such as ring, module, and crystal indices, because they are needed later for coincidence sorting and line of response construction.
When configuring singles for the PET scanner, check that the digitizer is connected to the correct hit collection, that the detector IDs match your ring and crystal geometry, and that the electronics time window for summing hits is realistic but small enough so that separate interactions are not merged by mistake. The output of this stage is a list of singles that can then be filtered in energy and combined into coincidences.
Digitizer singles represent processed detector events, not raw hits. Always verify that singles are built from the correct hit collection and that the detector IDs are consistent with your PET geometry.
Energy blurring
Real detectors do not measure the deposited energy perfectly. Instead, they show an energy resolution that broadens every monoenergetic line into a peak. To reproduce this effect in PET, you configure an energy blurring stage in the digitizer chain. This transforms the ideal energy from the singles into a smeared energy that mimics the detector response.
Energy resolution is usually specified as a full width at half maximum (FWHM) at a reference energy, often 511 keV for PET. A Gaussian smearing is then applied to the single energy, with a standard deviation $\sigma$ related to the FWHM by
$$\sigma = \frac{\text{FWHM}}{2.355}.$$
You choose a model for how the resolution scales with energy. A common simple model assumes that the relative resolution, $\text{FWHM}/E$, is constant. In practice, OpenGATE allows you to specify the resolution at a given energy, and it handles the conversion to a Gaussian width. During the digitization, each single’s energy is replaced by a random sample from this Gaussian distribution centered at the true deposited energy.
In the PET scanner project, you typically use energy blurring after singles creation, so that the summed energy in each crystal is blurred only once. When you define the blurring in the digitizer, you set parameters such as the reference energy, the resolution at that energy, and possibly coefficients to describe a more realistic energy dependence.
Energy blurring must be applied after energy summation to represent the detector channel response. Use the relation $\sigma = \text{FWHM} / 2.355$ to convert between FWHM and the Gaussian width used for smearing.
Energy window
After energy blurring, PET scanners apply an energy window to reject events that are unlikely to correspond to unscattered 511 keV photons. This selection is crucial to improve image quality because it reduces the contribution from scattered photons and background.
An energy window is defined by a lower and an upper threshold. For PET, the center of the window is typically around 511 keV. For example, a common choice is a 20 percent window such as 400 to 600 keV. Only singles whose blurred energy lies inside this range are kept. Those below the lower threshold are mostly heavily scattered photons or noise, while those above the upper threshold are rare and often due to pile-up or misreconstruction.
In the digitizer configuration, the energy window is implemented as a filter stage that reads the blurred energy and discards singles outside the specified bounds. The lower and upper limits must be defined in the correct units that match your simulation, usually keV. This selection is applied event by event and defines which singles will be used later by the coincidence sorter.
Choosing the window is a trade-off. A narrow window around the photopeak reduces scatter but also discards more true events, which lowers sensitivity. A wider window increases sensitivity but admits more scattered coincidences, which can degrade image contrast and quantitative accuracy.
The energy window is applied to the blurred energy, not the true deposited energy. A typical PET photopeak window is centered at 511 keV with a relative width of about 15 to 25 percent, for example 400 to 600 keV.
Coincidences
PET imaging is based on detecting pairs of photons from the same annihilation event that are registered in time coincidence. After singles are created, blurred, and filtered by the energy window, the digitizer must identify these pairs. This is the role of the coincidence sorting stage.
Coincidence sorting groups singles into pairs by comparing their detection times. If two singles occur within a specified coincidence time window, they are accepted as a coincidence event. This coincidence window is typically on the order of a few nanoseconds in clinical PET. In the digitizer configuration you define the size of this time window, and you also specify that only singles which have passed the energy window should be considered.
The coincidence sorter usually constructs one record per accepted pair, containing the energies, positions, times, and detector identifiers of the two singles. This information is sufficient to derive the line of response, to classify events as true, scattered, or random in analysis, and to build sinograms or list-mode data for reconstruction. The sorter can also exclude self coincidences where both singles come from the same detector channel, depending on your setup.
In the PET scanner project, configuring coincidences in the digitizer involves connecting the coincidence module to the singles output, setting the time window, and choosing what attributes are stored in the coincidence output. For time-of-flight PET, you keep accurate timestamps to allow later computation of the time difference and TOF position estimate, but the basic coincidence configuration is the same.
Coincidence sorting uses the detector timestamps and a coincidence time window to pair singles. Only singles that pass the energy window should enter the coincidence sorter, otherwise scattered and low energy events will contaminate the coincidence data.
Views: 11