KAHIBARO
32.4. Energy Windows
Table of Contents
Primary window
In SPECT acquisition, an energy window is a range of detected photon energies that you decide to keep for image formation. Most SPECT radionuclides emit one or more characteristic gamma rays with known energies, for example about 140 keV for Tc 99m. Ideally, you want to keep photons that underwent only photoelectric absorption in the crystal after minimal interaction in the patient or surrounding materials. These photons form the photopeak in the detector energy spectrum.
The primary window, often called the photopeak window, is centered on the main gamma energy of the radionuclide. Its purpose is to select events that are most likely to be unscattered primary photons. In GATE, this selection is usually implemented in the digitizer by defining a lower and upper energy threshold for accepted detector events. Only singles whose deposited energy lies inside this interval are kept as primary counts.
In a typical Tc 99m SPECT acquisition, once your energy blurring is configured to mimic realistic detector resolution, you inspect the measured energy spectrum from your simulation. The photopeak appears as a broadened peak around 140 keV. You then define a primary window around this peak, for example from 126 keV to 154 keV, which corresponds approximately to a 20 percent window symmetrically around the peak energy. Other radionuclides use different window widths, often between 10 and 20 percent of the photopeak energy, depending on detector energy resolution and clinical protocol.
Choosing the width of the primary window is a trade off. A narrow window reduces the fraction of scattered events but also loses some true primaries due to detector energy resolution. A wide window improves sensitivity because more counts are included, but it also admits a higher fraction of scattered photons. In SPECT, you often adjust this window width to match your intended image quality and noise level, and you keep it consistent between simulations if you want comparable results.
In practice, for SPECT acquisition in GATE, you first configure realistic energy blurring in the digitizer, so that your simulated spectrum resembles that of a real gamma camera. Then you use an energy window digitizer module, or equivalent selection, to specify the primary window. Events outside this energy interval are discarded from the projection data that you record with your projection actors. The primary window typically defines your main projection dataset that will be used for reconstruction.
When your SPECT acquisition involves multiple photopeaks, for example In 111 with gammas at about 171 keV and 245 keV, you can define one primary window per peak if you plan to reconstruct separate images. In that case, the energy window module in GATE can be used to label or split events into different output streams depending on which primary window they fall into.
Primary window selection rule: keep only events with detected energy within a chosen interval around the main gamma peak, typically defined as a percentage of the nominal photon energy, for example $E_0 \pm p \,\%$.
Scatter windows
Scatter windows are additional energy windows placed below the primary photopeak window. Their goal is to capture photons that have likely undergone Compton scattering in the patient or in surrounding materials. These events carry information about scatter contamination in the primary window and are commonly used for scatter correction methods in SPECT.
In a Tc 99m acquisition, the Compton continuum appears as a broad distribution of energies below the 140 keV peak. You define one or more scatter windows in a region of this continuum, for example a window just below the primary one, such as 120 keV to 126 keV, or a wider region further down. The exact placement and width of the scatter window depend on the scatter correction method you want to emulate, for example the dual energy window or triple energy window approaches.
In GATE, scatter windows are implemented in the same digitizer mechanism as the primary window, but with different energy limits and usually a different label or output collection. During acquisition, each detected event is assigned to the primary window, one of the scatter windows, or rejected, based on its detected energy after blurring. You can then record separate projection datasets for the primary and scatter windows using suitable actors. These datasets are later combined during data analysis to perform scatter correction, for example by subtracting a scaled version of the scatter projections from the primary projections.
Defining scatter windows also requires attention to realistic detector modeling. The position and shape of the Compton continuum depend on energy resolution, crystal thickness, and collimator scatter. If your energy blurring in GATE is not realistic, the fraction of counts migrating from the primary peak into the scatter window and vice versa will not match real systems. Before relying on scatter windows in your simulation, you typically validate the simulated energy spectrum against measured data for a simple phantom.
Scatter windows influence both image noise and bias. Wider scatter windows contain more counts, which can improve the statistical precision of scatter estimates, but they also include events with different scatter angles and origins. Narrow scatter windows are more specific but may be too noisy. In a simulation study, you can systematically vary scatter window definitions and quantify their impact on reconstructed image quality, which is one of the reasons to implement SPECT acquisition energy windows explicitly in GATE.
When you simulate radionuclides with multiple photopeaks, scatter windows may need to account for Compton scatter from higher energy photons that fall into the energy range of lower energy peaks. In such cases, you define scatter windows not only below each primary window but also in energy regions where down scattered higher energy photons contribute significantly.
Scatter window usage principle: define one or more windows below the photopeak to measure scatter dominated counts, and use the corresponding projections, usually scaled, to estimate and correct scatter in the primary window data.
Views: 14