The data acquired by the positron camera during the beam spills delivered by the heavy ion synchrotron SIS at GSI Darmstadt are affected by high noise level and are useless for the image reconstruction. A reason could be a wrong correction of random coincidences. The standard method of random correction, identifying random coincidences with delayed ones (delay time of 128 ns and coincidence window 12 ns wide), properly works in nuclear medicine, where the random background seen by the scanner is rather constant in time. In the therapy case, a background correlated in time with the beam microstructure on a time scale comparable to the delay time could lead to an underestimation of random coincidences and hence to the observed increase of noise.

Fig. 1. Model of the background radiation reaching the scanner during the beam spill.

In Fig. 1 we sketch the background model used in our simulation approach for reproducing the experimental random (i.e. not related to b⁺-activity) prompt and delayed coincidence rates acquired by the camera in spill. The periodical repetition of the bunch Gaussian shape represents the background radiation directly following the beam appearance, whereas the bias c introduces a delayed g-ray component, due to isomeric de-excitations or radioisotope decay with half-lives longer than several ns.
Our model with FWHM and period values given by in-ring measurements of the beam microstructure at flat top voltage U_ft = 2 kV (which is applied for a more homogeneous beam structure in time) could not explain all the experimental PET data. Therefore, the microstructure was measured after the extraction in the beam line of the medical cave. At U_ft = 2 kV the time occurrence of the ions in one RF period (Fig. 2) revealed bunches of lower FWHM than in the accelerator ring. With the new measured T and w values, all the experimental prompt and delayed coincidence rates could be reproduced by simulation, provided that a bias c of some percent with respect to the Gaussian modulation was added. However, at U_ft = 0 kV the microstructure vanished (Fig. 2), while the in spill PET data could be explained by a loss but not a completely suppression of time correlation at U_ft = 0 kV. Nevertheless, the disagreement could come from slight self-bunching effects, due to the higher beam intensity (I @  10⁷-10⁸ ions/s) in the PET acquisitions than in the microstructure investigation (I @  10⁵-10⁶ ions/s). On the basis of all these considerations, we want to measure the background radiation seen by the positron camera. If the experiment will confirm our model and hence prove the beam microstructure influence on the in spill PET acquisitions, the last step will be to destroy the microstructure or to find a new method of random correction, overcoming the limitations given by frozen software and hardware features of the scanner installed at GSI.

Fig. 2. Left: example of beam microstructure at U_ft = 2 kV. A Gaussian fit is superimposed onto the data. Right: example of not correlated time distribution measured at U_ft = 0 kV.

¹ Gesellschaft für Schwerionenforschung Darmstadt

 IKH 06/26/01 © K. Parodi