Ponente
Descripción
PET has consolidated as one of the most used and important molecular imaging techniques. However, there is still room for improvement. Specifically, boosting the system’s sensitivity will allow dose and/or time acquisition reduction without impacting the reconstructed image’s quality. BGO detectors have higher density when compared with the most used lutetium-based detectors, which translates into an increased sensitivity,. Moreover, the BGO crystals produce Cherenkov photons, generated much faster than the scintillation ones, that can be used and positively impact the timing performance of the PET detector. Furthermore, DOI capability is key in PET, for correcting parallax errors, allowing a uniform spatial resolution across the entire FOV.
We propose two BGO PET detectors designs, both with DOI and timing capabilities. The first is a semi-monolithic block of 1×8 slab of 3×25×15 mm3 each. All faces of each slab are polished and covered with ESR, except for the face coupled to the photosensor. The second detector proposed is a BGO crystal with the so-called pseudo-slabs geometry. This crystal is composed of 1×8 pseudo-slabs of 3×25×15 mm3 each, covered in BaSO4. Each pseudo-slab consists of 8 pixels of 3×3×15 mm3 with the four lateral faces unpolished and glued together. The entrance face of the block has an ESR film. The external dimensions of both crystal blocks is 25.8×25.8×15 mm3 and for all the studies these were coupled to a Silicon SiPM matrix. The TOFPET2 ASIC from PETsys Electronics was used as the readout electronics. Experimental data was acquired to study the spatial capabilities of the proposed detectors.
A uniform measurement was performed and the floodmap for each detector was generated. The x-monolithic and the y-pixelated direction were estimated using the RTP2 and CoG algorithms, respectively.
The monolithic spatial resolution of the semi-monolithic block and the DOI resolution of both crystals were studied. In each case a NN based on a MLP was used. For the training and testing along the monolithic direction, a slit collimator was moved in 1 mm steps between the detector under study and a single BGO pixel. For the DOI evaluation, the slit was displaced in 2 mm steps along the 25.8×15 mm2 face. In all cases the data was split into training, evaluation and testing datasets. The testing dataset was used for predicting the monolithic/DOI positions and the FWHM and MAE of the error profiles are reported. The mean monolithic spatial resolution of the semi-monolithic block was 3.4±1.1 mm. For the pseudo-slabs crystal, all pixels are well resolved. The mean DOI spatial resolution values was 3.8± 0.9 mm and 4.3±0.5 mm for the semi-monolithic and pseudo-slabs crystal, respectively.
The results obtained indicate that both crystals are suitable for a clinical PET system. We are currently working on the timing capabilities for these BGO blocks using ASICs.
