Aim/Introduction: Segmentation of lungs, tumor and normal liver is a major source of errors for 99mTc-MAA SPECT/CT-based dosimetry for 90Y radioembolization. The goal of this study is to assess different segmentation protocols for lung shunt fraction (LSF) and tumor-to-normal liver ratio (TNR) for 99mTc-MAA SPECT/CT. Materials and Methods: The XCAT phantom was used to simulate 10 patient anatomies with 99mTc-MAA distribution, with 5%-20% LSF, axial respiratory motion of 2 cm, different TNR and tumor size. An analytical projector for low energy high resolution parallel-hole collimator was used to simulate 128 realistic noisy projections. The OS-EM method was used for reconstruction with 4 iterations and 16 subsets with attenuation correction using averaged attenuation map, effective source scatter estimation and geometric collimator-detector-response modelling. For LSF, lungs were segmented based on different CT images, i.e., helical CT at end-inspiration (HCT-IN), helical CT at mid-respiration (HCT-MID), helical CT at end-expiration (HCT-EX) and cine averaged CT (CACT). For measuring the lung counts, 2.21 cm lung basal region was excluded to calculate the lung mean count density and then multiply with the original lung volume [1] from different CTs. Liver was segmented based on SPECT images with 2.5% lower threshold or contouring on respective CT images. For TNR, tumors and normal liver were segmented from intensity thresholding on SPECT, contouring on various contrast HCTs or by combining the information from both of them [2]. The mean errors of LSF and TNR for different phantoms were computed for different protocols. Results: For liver segmentation in various LSFs, the LSF errors are generally higher for CT-based segmentation as compared to those of SPECT-based segmentation, which can reach 15% when using HCT-EX to segment the lungs. The mean LSF error is lowest for the use of CACT to segment the lungs, i.e., <1%. The TNR errors can be over 30% when using contrast HCT-EX as the segmentation method. The mean TNR error is lowest for SPECT-based segmentation with 50% and 2.5% lower threshold for tumor and liver respectively. Conclusion: It is recommended to use SPECT thresholding based method to segment the liver and to use CACT to segment the lungs for LSF calculation. The TNR can be measured based on SPECT data with good quantitation accuracy without the use of contrast CT for 90Y radioembolization treatment planning. References: [1] Yu N, et al. Int J Radiat Oncol Biol Phys. 2013;85(3):834-9. [2] Zeintl J, et al. J Nucl Med. 2010;51:921-928.