Abstract
12
Objectives: We have developed a human brain PET insert that is compatible with 7T MRI scanners. The brain PET insert has a 16.7-cm-long axial field-of-view (FOV) that provides simultaneous 7T brain PET/MR images without a bed movement. Here, we present the scanner design and system performance of the brain PET insert in accordance with NEMA NU 2-2018 and NU 4-2008 standards.
Methods: Our brain PET insert is comprised of 18 sectors with a ring diameter of 33 cm. Each sector is encapsulated with a copper-based RF shielding box with non-magnetic components to operate within the 7T MRI scanners. A block detector consists of a dual-layer staggered DOI crystal array coupled with a 2×2 array of 16-channel Hamamatsu MPPCs. The upper layer is an 11×11 array of 2.09×2.09×8 mm3 LSO crystals and the lower layer is a 12×12 array of 2.09×2.09×12 mm3 LSO crystals. All the PET data were collected using an in-house developed FPGA DAQ system. The energy channels consist of gated charge-to-time converters (QTCs) combined with single-ended memory interface (SeMI) of FPGA, and the timing channels are based on 16-ps-resolution time-to-digital converters (TDCs) implemented within the same FPGA. The system performance was evaluated using an energy window of 350-650 keV and a fixed coincidence time-window of 4 ns, unless otherwise noted. The spatial resolution was measured based on the NEMA NU 4 standard using a 22Na point source embedded in a 1 cm3 acrylic cube. The point source was placed at the center of transverse FOV and sequentially stepped in a radial direction up to 100 mm with a 5-mm interval. The measurement was repeated at the axial center and 1/4-axial-offset position. The acquired data were reconstructed using both a 2D FBP and 3D OSEM. The sensitivity specified in the NEMA NU 4 standard was measured using the same 22Na point source. The point source was placed at the center of transverse FOV and axially stepped with a 2.17-mm interval to cover the entire axial FOV. The sensitivity specified in the NEMA NU 2 standard was measured using a [18F]-FDG line source with five aluminum tubes. The acquisition was repeated with the source axially shifted by 0 cm, 5 cm, and 10 cm from the isocenter of the PET scanner. The background count-rate was measured without locating the source and subtracted from each dataset. The sensitivity was measured at three different energy windows of 250-750 keV, 350-650 keV, and 450-550 keV. The count-rate performance specified in the NEMA NU 2 standard was measured using a 70-cm-long [18F]-FDG line source. The line source was inserted into a cylindrical polyethylene phantom and placed at the center of transverse FOV. The peak noise equivalent count rate (NECR) and scatter fraction were then estimated.
Results: The radial/tangential/axial resolution at the isocenter of the PET scanner was 2.53/2.51/2.28 mm FWHM (2D FBP) and 1.39/1.61/1.92 mm FWHM (3D OSEM). The spatial resolutions at each source position are summarized in Tables 1 (2D FBP) and 2 (3D OSEM). The sensitivity was 18.87/20.64/23.72 kcps/MBq at 0 cm, 5 cm, and 10 cm off-center (NEMA NU 2) and 6.19% (NEMA NU 4) at the energy window of 350-650 keV. The axial sensitivity profiles at three different energy windows are illustrated in Figures 1a (NEMA NU 2) and 1b (NEMA NU 4). The count-rate performance as a function of activity concentration was shown in Figure 2. The peak NECR was 11.9 kcps at 1.94 kBq/mL with the scatter fraction of 57%. An FPGA firmware upgrade is in progress to improve the count-rate performance.
Conclusions: In this study, we measured the system performance of our newly-developed 7T MR-compatible brain PET insert based on the NEMA standards. Our next milestone is to perform in-vivo human brain studies and assess the image quality using a cylindrical phantom that contains hot and cold spheres with six different diameters. Research Support: This work was supported by grants from National Research Foundation of Korea (NRF) funded by Korean Ministry of Science, ICT, and Future Planning (grant no. NRF-2014M3C7034000).