TY - JOUR T1 - Whole-body continuous-bed-motion PET list-mode reconstruction with non-rigid event-by-event respiratory motion correction JF - Journal of Nuclear Medicine JO - J Nucl Med SP - 105 LP - 105 VL - 60 IS - supplement 1 AU - Yihuan Lu AU - Kathryn Fontaine AU - Jean-Dominique Gallezot AU - Tim Mulnix AU - Vladimir Panin AU - Judson Jones AU - Michael Casey AU - Richard Carson AU - Chi Liu Y1 - 2019/05/01 UR - http://jnm.snmjournals.org/content/60/supplement_1/105.abstract N2 - 105Objectives: Recently, continuous-bed-motion (CBM) acquisition of PET was released by Siemens [1], which enhanced patient comfort [2] and demonstrated advantages in spatial uniformity and end-plane imaging as compared to the conventional step-and-shoot (SS) acquisition [3]. In the past, we have developed list-mode (LM) reconstruction equipped with respiratory event-by-event non-rigid motion correction (EBE-NRMC) [4] for single-bed PET that uses all counts. However, the existing EBE-NRMC method was not readily applicable to CBM due to the constantly-moving field-of-view during the acquisition. In this study, by taking the bed motion into account, we developed the first list-mode PET reconstruction with event-by-event non-rigid respiratory motion correction for whole-body (WB) CBM acquisition. Methods: During the CBM acquisition, bed position information, was encoded into the LM data stream. The bed motion can be viewed as rigid axial-direction patient motion. Therefore, the bed position information can be used to reposition the detector-pair coordinates of the line-of-response (LOR) for each event in the axial direction, to account for bed motion. Since normalization was computed for each LOR, there was no difference in CBM as compared to single-bed LM reconstruction [5]. The scatter correction was performed using 3D single-scatter-simulation technique with slice-by-slice calculated scaling factor in axial direction. CT-attenuation map was used for attenuation correction. Our approach reconstructs the entire WB image with only one LM reconstruction, whereas Siemens’ sinogram-based approach performs “chunk-by-chunk” reconstruction [1], in which the number of reconstructions is proportional to the length of a subject. To build the respiratory motion model, ANZAI-based respiratory-gated MLAA reconstruction was first performed [4], followed by inter-gate non-rigid image registrations. The internal-external correlation (INTEX) technique was then used to build a respiratory motion model, which describes a continuous relationship between each voxel’s movement (internal) and the ANZAI trace (external) [4]. The motion model, along with CT-attenuation map, list-mode data and bed motion information were then used in the final LM EBE-NRMC reconstruction, to achieve respiratory motion corrected reconstruction for whole-body CBM data. Evaluation was performed on four subjects, 2 healthy control (HC) and 2 lung cancer (LC) patients, who were scanned 60-90 min post 10 mCi 18F-FDG injection on Siemens Biograph mCT. A protocol of 6 WB passes at 5 min/pass was used. LM data from all passes (30 min) were reconstructed into one WB image. Organ-specific motion amplitude, and the % increase in SUV from uncorrected to EBE-NRMC corrected values were reported for each subject. Results: Compared with Siemens reconstruction, LM reconstruction showed comparable results in visual comparison (fig. 1(A-C)). Siemens recon appears to be smoother than LM recon due to different point-spread-functions were used. LM recon shows sharper brain structure (fig. 1(A,B)). Respiratory motion correction is effective, even for very small tumors with small motion amplitude, i.e. mean 140 mm3 in volume with <4mm motion (fig.1 (E, G)). SUV increase for large organs after respiratory motion correction is approximately proportional to the motion amplitude (fig.1 (E-G)). Conclusions: We have developed list-mode reconstruction for continuous-bed-motion acquisition, which is comparable to Siemens reconstruction without motion correction. Whole-body event-by-event non-rigid respiratory motion correction shows superiority in SUV and resolution recovery as compared to no motion correction. This approach has the potential to produce optimal WB static and dynamic images. References: [1] Panin et al., PMB, 2014. [2] Schatka et al., EJNMMI, 2016. [3] Osborne, et al., AJNMMI, 2015. [4] Lu et al., JNM, 2018. [5] Jin et al., PMB, 2013. ER -