Total body PET attenuation correction based on low dose CT and non-attenuation-corrected PET images

A total-body PET/CT scanner with a long axial field of view offers much sensitivity for clinical imaging and research. However, it may lead to more radiation exposure for the whole body rather than a single-site examination. In particular, more X-ray exposure is needed to cover the whole body instead of a single site as in common CT, which accounts for the major dose of radiation. Considering this motivation, we explored a dual dose reduction strategy to apply low-dose CT and low-dose PET scanning protocols on a total body system for PET attenuation correction. To conduct this dose reduction task, we proposed a deep learning method implemented with a novel network model equipped with several spatial position alignment modules, where multiscale feature information of low-dose CT images and low-dose NAC PET are fused and ultimately predict normal-dose PET images. Quantitative measurement metrics are used to evaluate the performance, and the experimental results prove the effectiveness of the proposed methods. Moreover, we found that 1.0% NAC PET could provide excellent identification of lesions, and 25% NAC PET scanning data generated more texture details than the ground truth. In the future, further research on dose reduction should be explored for total body PET/CT systems.