UQx Bioimg101x 6.7.1 PET Image Reconstruction


PET images are reconstructed in a similar fashion
to CT. The main difference is in how the signal is detected, in CT we project an external
beam through tissue and measure attenuation. In PET we detect the signal from an injected
radioactive isotope in the tissue. Here we see a typical PET image that has been overlaid
on a gray CT image. Note that the resolution is substantially less than that of CT or MRI,
this lower resolution is largely due to the smaller amount of information and lower signal
to noise that is available during acquisition. In order to understand PET reconstruction
we first need to review the process by which the PET signal is acquired. In a PET system
the subject is placed in the scanner. The scanners detector is called a scintillator
ring and can detect extremely low signals, in PET’s case this signal consists of a single
positron. In order to detect this each detector element is quite large. Recent advances in
technology have enabled smaller and smaller elements in this array which in turn leads
to better resolution. The use of cascade photo diodes to replace the scintillator ring also
increases the speed of acquisition that in turn will also increase resolution. In PET
imaging a radio tracer is then injected and once this binds with the target tissue a collision
event occurs, represented here as a star. This results in two positrons being emitted
in exactly opposite directions to each other at the speed of light. Given their velocity,
irrespective of their location in the object these two positrons will hit the detector
ring at essentially the same time. When this happens the scanner records what is called
a co-incidence event. As more of the PET tracer binds at a location more collision events
are then recorded as time progresses. The time scale over which this happens is dependant
on the tracers binding potential and the abundance of what is being bound but typically is between
6 and 12 nanoseconds. Note that not all collision events will be recorded, only those that occur
in the same plane as the scintillator ring. This is another factor that reduces the amount
of signal that is available in a PET scanner. Note that some modern scanners can detect
some of the out of plane coincidence events but only a small percentage more. If we now
look at the case of where multiple simultaneous tracer bindings are occurring. We can see
that there will be multiple over-lapping co-incidence events. While the events themselves are occurring
simultaneously, the detection of the co-incidence events will still occur independently given
the immense speed of the positrons. The raw output data from the scanner is then a series
of points that represent the co-incidence events, each event is represented as two points
through which a line can be drawn. This is called a Line or Response or LOR. From the
data we don’t know where along the LOR the event occurred. More modern detectors with
a higher time resolution can give an approximation of where the event occurred within an accuracy
of about a hundred millimetres. These systems are called Time of Flight or TOF detectors
and operate on a time scale of approximately three nanoseconds. We can now visualise the
reconstruction process by placing a probabilistic line between each of the two points from the
co-incidence event raw data. As we add more and more of these events. The result is a
final probabilistic map of the source of the events. We haven’t displayed the exact technical
details of how this occurs as there are many differing techniques. The most typically used
technique involves first sorting the LOR’s into similar directions, grouping them into
what are called sinograms and then reconstructing them as per CT backprojection. Other techniques
used statistical techniques such as expectation maximisation to generate an inverse solution.
What is similar between all these techniques is that the end result is a probabilistic
map of the source of the co-incidence events.

You may also like...

1 Response

  1. Juhan says:

    It's not he positrons that are detected, but the two same energy (0.511 MeV) photons that are produced in the process of annihilation between the positron and an electron in the tissue.

Leave a Reply

Your email address will not be published. Required fields are marked *