Aim/Introduction: This study analysed different workflows to compare dosimetry computed from SPECT99m Tc-MAA images and PET treatment monitoring images following 90Y-radioembolisation. Materials and Methods: A cohort of 20 patients iscurrently being studied. For each patient, 99m-Tc-MAA SPECT/CT images were acquired, reconstructed with attenuation- and diffusion-corrections, and calibrated in MBq.mL-1. Treatment monitoring was performed with a digital PET/CT using reconstruction parameters [1] optimised with dose-volume histograms (DVHs). The whole liver (WL) and total tumour (T total) were contoured on the CT acquired during the workup session and propagated to the SPECT images. The perfused liver volume (PLV) was defined using a 5% threshold on the SPECT. Contours were propagated to PET images with four registration methods, rigid global, deformable global, rigid local and rigid local,functional. The two firsts consider CT (SPECT) to CT (PET) rigid/deformable (BSpline) transformations. The third uses a rigid transformation between the two CTs, centred on the live ronly. The last one considers rigid local registration between SPECT and PET images with a rigid local transformation centred on the liver. Absorbed dose distributions were calculated using Voxel S-values convolution. The predictive and treatment follow-up DVHs were compared using the mean absorbed dose (D mean) and absorbed dose at 2% volume (D 2%). Results: For the first patients, D mean were within 5.4% and 6.4% and D2% were within 2.0% and 2.1% for the WL and PLV whatever the registration method, respectively. However, for Ttotal, the range was larger. For example, for patient #3, the SPECT-predicted Dmean was125.4 Gy while it was 39.07, 43.6, 30.7 and 168.3 Gy from the PET images for all methods, respectively. In that case, only locally rigid registration between functional images (SPECT and PET) instead of CT images provided comparable DVHs. This might be due to inaccurate location of the lesions due to respiratory motion and internal organ movements between work up and treatment sessions. Even if liver CT contours were well registered, the activity measured from SPECT and PET, integrated over a longer acquisition duration than CT (12.5 min/bed for SPECT, 10 min/bed for PET) mismatche mismatched. Conclusion: rigid local,functional method is recommended in addition to other CT registration methods to propagate contours from SPECT/CT to PET/CT. The analysis on the entire cohort will be presented at the conference. References: [1] Labour, J. et al. "Yttrium-90Quantitative Phantom Study Using Digital Photon Counting PET." EJNMMI Physics, in revision, (2021).