Evaluation of dual-ended readout GAGG-based DOI-PET detectors with different surface treatments

Parallax error is a major issue in small animal positron emission tomography (PET) scanners which are used in preclinical studies or detailed scanning of human organs. Several methods have been proposed and investigated to reduce this radial artifact in PET images by estimating the depth of interaction (DOI) of 511-keV photons in the crystal. Among all, the dual-ended readout method seems to be very simple and effective as it does not have any fabrication and readout complications. In the past, some studies suggested that increasing the roughness of crystal lateral surfaces improves DOI resolution. In this paper, this was experimentally examined for four Ce:GAGG crystals with different surface structures.

Four 1.2 × 1.2 × 20 mm³ GAGG crystals with following surface treatment were examined: polished with optical finishing, fine grinding (using a fine surface grinding machine), fine cutting (no treatment), and coarse cutting (no treatment). These crystals were coupled individually to two SiPMs for dual-ended readout and placed in a coincidence detection circuit for electronic collimation of 511 keV incidents. The crystals were compared in terms of energy response and DOI estimation capability.

DOI function for each crystal was extracted and FWHM DOI resolution was calculated. DOI resolution for the polished crystal varied in the range of 0.54-4.14 mm throughout the length of the crystal due to its nonlinear DOI function. The fine grinding crystal showed a linear DOI function within the dynamic range of (-0.75, 0.75), and its DOI resolution varied in the range of 1.24-1.50 mm (1.37 ± 0.13 mm DOI resolution). The fine-cut crystal had almost a linear DOI function and a wider dynamic range of (-0.85, 0.85) and therefore the best performance with 1.2 ± 0.08 mm DOI resolution. However, for the crystal with the roughest surface (coarse-cut crystal), even though the dynamic range expanded to (-0.95, 0.95), its DOI function became nonlinear resulting in 1.24 ± 0.28 mm DOI resolution. This means there is an optimum surface roughness to provide the crystal with the best DOI capability. The pulse-height spectrum measured at each depth was used as a measure to compare the energy performance of the four crystals. The photopeak of 511 keV was observed for all depths, all crystals. The photopeak position for the coarse-cut crystal had extensive depth dependency which results in poorer energy resolution unless the energy window is calibrated for each depth. This variation of photopeak for the fine-cut and fine grinding crystals was comparable with that of polished crystal.

This paper reports 1.2 ± 0.08 mm FWHM DOI resolution for a fine-cut unpolished crystal. This resolution is as narrow as the crystal width, resulting in the complete elimination of parallax error in PET images. Results suggest that there is an optimum roughness for the best performance of the dual-ended method and further increase in the roughness, degrades DOI resolution. Thanks to the high light yield of GAGG, the energy performance of the fine-cut crystal is acceptable, and the depth dependency of the spectrum is negligible.