Development of a high resolution and quantitative SPECT for the human brain

We are developing a high resolution and quantitative SPECT for the human brain. The SPECT has two types of detector, a large FOV detector with a parallel collimator and a small FOV detector with a pinhole collimator. A quantitative and high resolution image is reconstructed by the small FOV detector using a supporting image obtained by the large FOV detector. The large FOV detector consists of a Nal(Tl) scintillator (147 × 250 × 6.4 mm 3 ) and 15 flat panel type multi-anode PMTs (H8500, Hamamatsu) arranged on 5×3 matrix. H8500 has 8×8 anodes with 5.8 mm 2 . We have constructed the large FOV detector and checked the performance. We also took SPECT images of a cylindrical phantom, a 3D brain phantom and multi line sources. An interaction point is calculated by the Anger method using all anodes (40×24). The observed intrinsic spatial resolution (FWHM mm) in x and j-direction were 3.5 mm and 3.1 mm, respectively. But these are less than we expected. We aimed at ∼2 mm, which is better resolution than clinical SPECTs ∼4, 5 mm. To improve the spatial resolution, we proposed a method instead of the Anger method and estimated the performance by using of a Monte Carlo simulation of scintillation lights (Geant4). To identify an interaction point, reference data set of distribution of scintillation lights on the PMT anodes is used. The reference data set is previously measured at known incident positions. For each event, the most similar distribution with the reference data set determined by a least squared method is identified as the interaction point. Using the method, 1.8 mm resolution is expected. Meanwhile, the small FOV detector is under the construction. We use a LaBr 3 (Ce) scintillator (100 × 100 × 4 mm 3 ) and 4 H8500s. We measured the physical performance, spatial and energy resolution and the position dependence. The intrinsic spatial resolution in × and j-direction and the energy resolution are 2.3 mm, 2.4 mm and 6.1 %, respectively.