Accurate and precise TOF-PET brain imaging despite inevitable system modelling errors

Positron emission tomography (PET) is a molecular imaging technique. A tracer molecule labeled with a positron emitter is administered to the patient. Upon decay, the positron annihilates with an electron and two photons are emitted back to back. When both photons are detected, we know that there is a tracer molecule somewhere along the line between the two detector elements. From a set of such PET measurements a tracer concentration image is reconstructed.
Recent PET systems also measure the difference in arrival time of the photons (time of flight, TOF) to estimate where along the line the positron was annihilated. TOF-PET data provide much more information than non-TOF data, enabling us to not only compute the tracer concentration, but also other information, such as the distribution of the attenuating material in the field of view. This enables quantitative PET without the help of an anatomical scan (CT or MR).
In ongoing work, we find that images computed with new algorithms exploiting TOF information do often not agree well with conventional images, although theory says they should. We have shown that this is due to inaccuracies in the current system calibration, which creates artefacts in the new as well as in the conventional images.
In this project, we will develop new algorithms for improved system calibration, where most of this calibration is done from the clinical TOF-PET data themselves, making the imaging more accurate, precise and robust than it is today.