Calibration of PET systems

There is a significant difference between human PET devices and small animal systems. This affects the production of the detector, the corrections preceding reconstruction and the corrections that are to be integrated in reconstruction. The table below contains the most important features.

feature device	small animal	human
diameter	15-20 cm	70-90 cm
non- collinearity	not significant	significant
crystal cross section	1.2-1.5 mm	3-6 mm
DOI	significant	not significant
crystal length	9-14 mm	20-24 mm
scattering (within body)	not significant	significant
scattering (within detector)	significant	not significant
energy resolution	important	very important
AFOV	5-12 cm	15-20 cm
length of examined object	8-15 cm	1.6-2m
random coincidence	important	very important
attenuation correction	useful	very important
ToF	not possible	possible
temporal resolution	1.2-4 ns	0.5(ToF)-7 ns
spatial resolution	1,2-2,5 mm	2,5-7 mm
PSPMT	almost always	never

 
The calibration procedure of the PET system is highly dependent on the device itself as well as on its process chain. Moreover, the order of steps can also change, but most often it includes the following main processes:

• Gain calibration after the ‘warm-up’ of the PMTs, measuring preferably in working coincidence mode, because it allows for performing the calibration of the amplification according to the position of the 511 keV photopeak.
• Delay (cable length) calibration. This also needs working coincidences and/or a reference detector. The timing of the PMTs in the module has to be synchronized based on measurement data containing time.
• Positioning calibration. It needs a flood-field image with good statistics.
• Timing differences between modules (offset). The time spectra of the module pairs have to be centred in an operating ring with the aid of a measurement containing time data, usually not by cables but by digital delay.
• Energy distribution map by crystal needles. It is possible only after performing the positioning calibration, the energy spectra have to be assorted by crystal needles, the 511 keV peaks of the needle by needle energy spectra have to be scaled to each other.
• Needle by needle timing calibration, practically only in ToF PETs; the local changes of the light yield of individual needles can cause different extents of delay, which can be corrected.
• Sensitivity calibration or normalization: the sensitivity of the individual crystal needles is never identical due to the inhomogeneous light yield and electronic gain, thus the sensitivity of the LORs – which approximately depends on the product of the sensitivity of two crystals – has a high standard deviation. This can be taken into consideration by using a sensitivity correction factor. (Clever averaging is important for obtaining good statistics; Defrise’s fan method is the most widespread.)

 

---