부산대학교 도서관

A high-resolution CT (HRCT) used as anatomical prior information during PET reconstruction can enhance the quality of a corresponding low-resolution PET image, provided that it is accurately registered to the PET dataset of interest. In this work, the impact of different PET/CT attenuation correction (AC) protocols on the registration between a dual-gated cardiac $^{18}{\rm F}$-FDG PET image and an HRCT image is investigated. The aim is to explore the impact of AC on PET-to-HRCT registration, and to identify the AC strategy that yields the best alignment between the left-ventricles in the PET and the HRCT images for subsequent partial volume correction. Simulations were performed using XCAT phantoms. Shallow breathing and a regular beating pattern were simulated and both noise-free and noisy data were evaluated. Respiratory motion during the acquisition of the CT used for attenuation correction strongly affected the dual-gated PET reconstructions, resulting in artefacts and quantification errors in the PET image and poor PET-to-HRCT registration accuracy. The blurring introduced by the beating heart, on the other hand, proved to have a negligible effect on PET-CT registration. Dual-gated PET images reconstructed without attenuation correction could be well registered to the HRCT if a good initial alignment between the starting images was provided. A commercially available strategy to deal with an AC CT that is acquired in the wrong respiratory phase was also evaluated, and yielded not only enhanced quantitative accuracy but also accurate PET-to-HRCT registration. The effect of a high level of noise, as present in a dual-gated cardiac PET study, was also investigated. Registrations proved to be sensitive to noise, but noise is not a major limiting factor for PET-to-HRCT registration. A selection of the investigated attenuation correction procedures was also evaluated using cardiac PET/CT data measured in sheep. The PET-to-HRCT registration performance confirmed the XCAT-based predictions.