Introduction to PET Physics

[Contents] [Section 1] [Section 2] [Section 3] [Section 4] [Section 5] [Section 6]

3. 2D mode and 3D mode

3.1 Principles of operation
3.2 Sensitivity to true coincidence events
3.3 Sensitivity to scattered events
3.4 Sensitivity to random events
3.5 Effect of camera geometry


3.1 Principles of operation

Most cameras employing block-detector technology (section 5.6) may be operated either in "2D" mode or "3D" mode. In 2D mode thin septa of lead or tungsten separate each crystal ring and coincidences are only recorded between detectors within the same ring or lying in closely neighbouring rings. Coincidences between detectors in closely neighbouring rings are summed (figure 6) or rebinned to produce a dataset consisting of 2P+1 co-planar sets of LORs normal to the axis of the camera, where P is the number of detector rings. Such a dataset may be reconstructed into images using standard tomographic techniques (section 4).


Figure 6. Axial cut-away view of a multi-ring PET camera (not to scale) operating in 2D mode, showing direct and cross-plane rebinning. In direct-plane rebinning (top), data in corresponding LORs with ring difference 0 and +/-2 are summed, whereas in cross-plane rebinning (bottom) corresponding LORs with ring difference +/-1 and +/-3 are summed. More or less LORs may be summed depending on the camera geometry and the imaging requirements.

In 3D mode, the septa are removed, and coincidences are recorded between detectors lying in any ring combination (figure 7). The rebinning technique described above may be directly applied to this 3D dataset (Daube-Witherspoon and Muehllehner 1987), but this results in significant image distortion which increases towards the edge of the FOV. Usually more computationally intensive fully-3D image reconstruction techniques are employed. The computational burden increases with the number of crystal rings used, and for cameras with a large number of rings some degree of rebinning between closely neighbouring LORs may be applied to reduce the dataset to more manageable proportions (e.g. Brix et al 1997) - this process is known as 'mashing'.

[top of page]

3.2 Sensitivity to true coincidence events.

Removal of the septa allows the use of a much larger number of measured LORs (see figure 7). This results in a significant increase in sensitivity to true coincidence events (Cherry et al, 1991). The increase in the number of LORs depends on the number of crystal rings and the degree of rebinning used in 2D mode. Rebinning in 2D mode results in a variation in sensitivity along the axial FOV. In 3D mode there is a much stronger variation in sensitivity (see figure 8), which peaks in the centre of the axial FOV.


Figure 7. Axial cut-away view of a PET camera in 2D and 3D mode showing how the number of possible LORs can increase when the septa are removed. For a 16-ring camera where the maximum permissible ring difference in 2D mode is +/-3, just over 2.5 times as many LORs are used in 3D mode than in 2D mode.

The septa themselves subtend an appreciable solid angle in the camera FOV, and as a result they have a significant shadowing effect on the detectors. The magnitude of the septa shadowing effect can be as high as 50%, and this is another reason why 3D mode is more sensitive than 2D mode. The increase in sensitivity to true coincidences is the prime motivation for the removal of septa, although the increased cost of manufacturing cameras with septa is also a factor.

Figure 8. Predicted sensitivity from the number of LORs used in 2D and 3D mode. The sensitivity is plotted as a function of image plane number for a 16-ring camera where the maximum ring difference in 2D mode is +/-3. At the edge of the axial FOV 2D mode and 3D mode have the same predicted sensitivity but in the centre of the FOV the sensitivity in 3D mode is significantly greater.

[top of page]

3.3 Sensitivity to scattered events.

In the presence of septa, only those photons scattered in the plane of each detector ring have a reasonable probability of being detected. When the septa are removed, it is possible to detect photons with a much greater range of scattering angles (figure 9). As a result, the sensitivity to scattered events in 3D mode increases significantly (Cherry et al,1991).

Figure 9. Effect of septa removal on sensitivity to scattered coincidences

[top of page]

3.4 Sensitivity to random events.

When the septa are removed, the FOV for single events is increased (figure 10). This can result in a significant increase in the number of random coincidences detected, particularly when imaging near organs which may contain significant amounts of activity, such as the brain, heart or bladder.


Figure 10. Effect of septa removal on sensitivity to single events

[top of page]

3.5 Effect of camera geometry.

The geometry of the camera also has a significant effect on performance, particularly in 3D mode, where increasing the axial FOV or the camera acceptance angle not only increases the sensitivity to true coincidences, but also increases the sensitivity to randoms and scatter. While the sensitivity to randoms can be reduced by increasing the radial extent of the end-shields (Spinks et al 1998), it remains a serious problem in whole-body imaging and for cameras using large-area planar detectors.

The most commonly used large-area planar systems employ two opposing Anger cameras (see section 5.2) and are known as dual-headed coincidence imagers (DHCI). Such systems are intended to be used both as PET and as SPECT cameras. One of the requirements for SPECT is that the detectors are close to the patient - unfortunately this increases the acceptance angle for single events when operating in PET mode. Some manufacturers attempt to address this problem by using coarse axial collimation, a configuration which has been described as "2.5D".


[top of page]

Last revised by:

Ramsey Badawi

Revision date:

12 Jan 1999