PET/CT for Brain Amyloid: A Feasibility Study for Scan Time Reduction by Deep Learning.

PURPOSE: This study was to develop a convolutional neural network (CNN) model with a residual learning framework to predict the full-time 18F-florbetaben (18F-FBB) PET/CT images from corresponding short-time scans. METHODS: In this retrospective study, we enrolled 22 cognitively normal subjects, 20 patients with mild cognitive impairment, and 42 patients with Alzheimer disease. Twenty minutes of list-mode PET/CT data were acquired and reconstructed as the ground-truth images. The short-time scans were made in either 1, 2, 3, 4, or 5 minutes. The CNN with a residual learning framework was implemented to predict the ground-truth images of 18F-FBB PET/CT using short-time scans with either a single-slice or a 3-slice input layer. Model performance was evaluated by quantitative and qualitative analyses. Additionally, we quantified the amyloid load in the ground-truth and predicted images using the SUV ratio. RESULTS: On quantitative analyses, with increasing scan time, the normalized root-mean-squared error and the SUV ratio differences between predicted and ground-truth images gradually decreased, and the peak signal-to-noise ratio increased. On qualitative analysis, the predicted images from the 3-slice CNN model showed better image quality than those from the single-slice model. The 3-slice CNN model with a short-time scan of at least 2 minutes achieved comparable, quantitative prediction of full-time 18F-FBB PET/CT images, with adequate to excellent image quality. CONCLUSIONS: The 3-slice CNN model with a residual learning framework is promising for the prediction of full-time 18F-FBB PET/CT images from short-time scans.

Development of PET/MRI with insertable PET for simultaneous PET and MR imaging of human brain.

PURPOSE: The purpose of this study was to develop a dual-modality positron emission tomography (PET)/magnetic resonance imaging (MRI) with insertable PET for simultaneous PET and MR imaging of the human brain. METHODS: The PET detector block was composed of a 4 × 4 matrix of detector modules, each consisting of a 4 × 4 array LYSO coupled to a 4 × 4 Geiger-mode avalanche photodiode (GAPD) array. The PET insert consisted of 18 detector blocks, circularly mounted on a custom-made plastic base to form a ring with an inner diameter of 390 mm and axial length of 60 mm. The PET gantry was shielded with gold-plated conductive fabric tapes with a thickness of 0.1 mm. The charge signals of PET detector transferred via 4 m long flat cables were fed into the position decoder circuit. The flat cables were shielded with a mesh-type aluminum sheet with a thickness of 0.24 mm. The position decoder circuit and field programmable gate array-embedded DAQ modules were enclosed in an aluminum box with a thickness of 10 mm and located at the rear of the MR bore inside the MRI room. A 3-T human MRI system with a Larmor frequency of 123.7 MHz and inner bore diameter of 60 cm was used as the PET/MRI hybrid system. A custom-made radio frequency (RF) coil with an inner diameter of 25 cm was fabricated. The PET was positioned between gradient and the RF coils. PET performance was measured outside and inside the MRI scanner using echo planar imaging, spin echo, turbo spin echo, and gradient echo sequences. MRI performance was also evaluated with and without the PET insert. The stability of the newly developed PET insert was evaluated and simultaneous PET and MR images of a brain phantom were acquired. RESULTS: No significant degradation of the PET performance caused by MR was observed when the PET was operated using various MR imaging sequences. The signal-to-noise ratio of MR images was slightly degraded due to the PET insert installed inside the MR bore while the homogeneity was maintained. The change of gain of the 256 GAPD/scintillator elements of a detector block was <3% for 60 min, and simultaneous PET and MR images of a brain phantom were successfully acquired. CONCLUSIONS: Experimental results indicate that a compact and lightweight PET insert for hybrid PET/MRI can be developed using GAPD arrays and charge signal transmission method proposed in this study without significant interference.

A prototype MR insertable brain PET using tileable GAPD arrays.

PURPOSE: The aim of this study was to develop a prototype magnetic resonance (MR)-compatible positron emission tomography (PET) that can be inserted into a MR imager and that allows simultaneous PET and MR imaging of the human brain. This paper reports the initial results of the authors' prototype brain PET system operating within a 3-T magnetic resonance imaging (MRI) system using newly developed Geiger-mode avalanche photodiode (GAPD)-based PET detectors, long flexible flat cables, position decoder circuit with high multiplexing ratio, and digital signal processing with field programmable gate array-based analog to digital converter boards. METHODS: A brain PET with 72 detector modules arranged in a ring was constructed and mounted in a 3-T MRI. Each PET module was composed of cerium-doped lutetium yttrium orthosilicate (LYSO) crystals coupled to a tileable GAPD. The GAPD output charge signals were transferred to preamplifiers using 3 m long flat cables. The LYSO and GAPD were located inside the MR bore and all electronics were positioned outside the MR bore. The PET detector performance was investigated both outside and inside the MRI, and MR image quality was evaluated with and without the PET system. RESULTS: The performance of the PET detector when operated inside the MRI during MR image acquisition showed no significant change in energy resolution and count rates, except for a slight degradation in timing resolution with an increase from 4.2 to 4.6 ns. Simultaneous PET/MR images of a hot-rod and Hoffman brain phantom were acquired in a 3-T MRI. Rods down to a diameter of 3.5 mm were resolved in the hot-rod PET image. The activity distribution patterns between the white and gray matter in the Hoffman brain phantom were well imaged. The hot-rod and Hoffman brain phantoms on the simultaneously acquired MR images obtained with standard sequences were observed without any noticeable artifacts, although MR image quality requires some improvement. CONCLUSIONS: These results demonstrate that the simultaneous acquisition of PET and MR images is feasible using the MR insertable PET developed in this study.

Development of brain PET using GAPD arrays.

PURPOSE: In recent times, there has been great interest in the use of Geiger-mode avalanche photodiodes (GAPDs) as scintillator readout in positron emission tomography (PET) detectors because of their advantages, such as high gain, compact size, low power consumption, and magnetic field insensitivity. The purpose of this study was to develop a novel PET system based on GAPD arrays for brain imaging. METHODS: The PET consisted of 72 detector modules arranged in a ring of 330 mm diameter. Each PET module was composed of a 4 × 4 matrix of 3 × 3 × 20 mm(3) cerium-doped lutetium yttrium orthosilicate (LYSO) crystals coupled with a 4 × 4 array three-side tileable GAPD. The signals from each PET module were fed into preamplifiers using a 3 m long flat cable and then sent to a position decoder circuit (PDC), which output a digital address and an analog pulse of the interacted channel among 64 preamplifier signals transmitted from four PET detector modules. The PDC outputs were fed into field programmable gate array (FPGA)-embedded data acquisition (DAQ) boards. The analog signal was then digitized, and arrival time and energy of the signal were calculated and stored. RESULTS: The energy and coincidence timing resolutions measured for 511 keV gamma rays were 18.4 ± 3.1% and 2.6 ns, respectively. The transaxial spatial resolution and sensitivity in the center of field of view (FOV) were 3.1 mm and 0.32% cps/Bq, respectively. The rods down to a diameter of 2.5 mm were resolved in a hot-rod phantom image, and activity distribution patterns between the white and gray matters in the Hoffman brain phantom were well imaged. CONCLUSIONS: Experimental results indicate that a PET system can be developed using GAPD arrays and the GAPD-based PET system can provide high-quality PET imaging.