Direct comparison between SiPM- and PMT-PET/CT in [¹⁸F]FDG brain imaging

Purpose: New PET detector systems have silicon photomultipliers (SiPM) instead of photomultiplier tubes (PMT). Positron emission tomography (PET) images acquired using SiPM results in less noise and higher contrast than PMT-PET and improves the detectability and quantifiability of small lesions in oncological PET images. Whether SiPM-PET is useful for brain imaging is unclear. The present study aimed to clarify the advantages of SiPM-PET/CT in [¹⁸F]FDG brain images in head-to-head comparisons between two PET scanners in phantom and clinical studies.

Methods: Images of Hoffman (Data Spectrum Corporation) and pool (Itoi Factory Inc.) phantoms containing [¹⁸F]FDG (20 MBq) were acquired for 30 min using Discovery MI (DMI; GE Healthcare) and Discovery PET/CT 710 (D710; GE Healthcare) scanners. Thereafter, PET images were reconstructed under clinical conditions. The ratio (%) of grey-to-white matter contrast (%contrast) and image noise (coefficient of variation, CV; %) and uniformity (standard deviation, SD) were respectively evaluated from images acquired from the Hoffman and pool phantoms. We administered 155.8 ± 14.7 MBq of FDG to 22 participants, then sequential acquisition of images were started at 40 min post injection, for 10 min using the D710 and the DMI. The mean (± SD) of interval of time between sequential acquisitions (D710 to DMI) was 15.2 ± 1.0 min. All [¹⁸F]FDG images were separately normalized to a standard [¹⁸F]FDG PET template using MIMneuro (MIM Software Inc.). Anatomical volumes of interest (VOI) were automatically placed on the caudate nucleus, cerebellum, the frontal, occipital, parietal and temporal lobes, the putamen, thalamus and whole brain. Mean standardized uptake values (SUV_mean) were measured using anatomical VOI. Statistical analysis of SUV_mean for both acquisitions was performed with a 2-tailed paired Student t-test. Linear Regression analysis was used to evaluate the relationship between the change of SUV_mean in whole brain and the interval between acquisitions. Values with P < 0.05 were considered significant.

Results: In the phantom studies, the contrast, image noise, and uniformity of DMI and D710 were 72.3% and 52.4%, 10.9% and 11.1%, and 0.0157 and 0.0373, respectively. All indices of the DMI satisfied the Japanese Society of Nuclear Medicine image quality acceptance criteria (%contrast > 55%; CV &#8804; 15%; SD &#8804; 0.0249). In the human brain studies, the SUV_mean in all region of DMI were increased significantly than those of the D710. The mean ± SD of the SUV_mean of whole brain were 6.1 ± 0.8 and 5.5 ± 0.8 for the DMI and D710, respectively. The R² of SUV_mean were 0.04 (P = 0.37). The SUV_mean was independent of the interval between acquisitions.

Conclusions: Improved contrast, uniformity and noise on brain FDG images acquired using SiPM-PET contributed to higher spatial and timing resolution. Therefore, SiPM-PET offers better quality and more accurate quantity PET brain images.