Position emission tomography (PET) is widely used in clinics and research due to its quantitative merits and high sensitivity, but suffers from low signal-to-noise ratio (SNR). Recently convolutional neural networks (CNNs) have been widely used to improve PET image quality. Though successful and efficient in local feature extraction, CNN cannot capture long-range dependencies well due to its limited receptive field. Global multi-head self-attention (MSA) is a popular approach to capture long-range information. However, the calculation of global MSA for 3D images has high computational costs. In this work, we proposed an efficient spatial and channel-wise encoder-decoder transformer, Spach Transformer, that can leverage spatial and channel information based on local and global MSAs. Experiments based on datasets of different PET tracers, i.e., 18 F-FDG, 18 F-ACBC, 18 F-DCFPyL, and 68 Ga-DOTATATE, were conducted to evaluate the proposed framework. Quantitative results show that the proposed Spach Transformer framework outperforms state-of-the-art deep learning architectures.
The Journal of Medical Imaging allows for the peer-reviewed communication and archiving of fundamental and translational research, as well as applications, focused on medical imaging, a field that continues to benefit from technological improvements and yield biomedical advancements in the early detection, diagnostics, and therapy of disease as well as in the understanding of normal conditions.
Current guidelines for administered activity (AA) in pediatric nuclear medicine imaging studies are based on a 2016 harmonization of the 2010 North American Consensus guidelines and the 2007 European Association of Nuclear Medicine pediatric dosage card. These guidelines assign AA scaled to patient body mass, with further constraints on maximum and minimum values of radiopharmaceutical activity. These guidelines, however, are not formulated based upon a rigor-ous evaluation of diagnostic image quality. In a recent study of the renal cortex imaging agent 99mTc-DMSA (Li Y et al 2019), body mass-based dosing guidelines were shown to not give the same level of image quality for patients of differing body mass. Their data suggest that patient girth at the level of the kidneys may be a better morphometric parameter to consider when selecting AA for renal nuclear medicine imaging. The objective of the present work was thus to develop a dedicated series of computational phantoms to support image quality and organ dose studies in pediatric renal imaging using 99mTc-DMSA or 99mTc-MAG3. The final library consists of 50 male and female phantoms of ages 0 to 15 years, with percentile variations (5th to 95th) in waist circumference (WC) at each age. For each phantom, nominal values of kidney volume, length, and depth were incorporated into the phantom design. Organ absorbed doses, detriment-weighted doses, and stochastic risks were assessed using ICRP reference biokinetic models for both agents. In Monte Carlo radiation transport simulations, organ doses for these agents yielded detriment-weighted dose coefficients (mSv/MBq) that were in general larger than current ICRP values of the effective dose coefficients (age and WC-averaged ratios of eDW/e were 1.40 for the male phantoms and 1.49 for the female phantoms). Values of risk index (ratio of radiation-induced to natural background cancer incidence risk x 100) varied between 0.062 (newborns) to 0.108 (15-year-olds) for 99mTc-DMSA and between 0.026 (newborns) to 0.122 (15-year-olds) for 99mTc-MAG3. Using tallies of photon exit fluence as a rough surrogate for uniform image quality, our study demonstrated that through body region-of-interest optimization of AA, there is the potential for further dose and risk reductions of between factors of 1.5 to 3.0 beyond simple weight-based dosing guidance.