Objectives: Yttrium-90 (90Y) beta minus emission allows a high and well located dose delivery in the treatment of unresectable hepatocellular carcinomas, but the absence of low energy gamma emission limits accurate dosimetry with SPECT. Most investigators use bremsstrahlung photons to generate SPECT images of 90Y biodistribution. As an alternative, the use of PET has recently been proposed, detecting the annihilation photons generated after internal pair production in 32 out of 10 millions decays. The aim of this study is to compare the effectiveness of PET and SPECT for the dosimetry of 90Y microspheres therapies. Methods: A Philips Gemini scanner, energy window 410-660 keV, was used for PET. A Prism 3000XP scanner with MEGP collimators, 50-150 keV window, was used for SPECT. Quantitative accuracy was compared with an anthropomorphic torso phantom. A liver:lung ratio of 5.25:1 was chosen to represent an average patient. Total activity was 370 MBq. Scan times were prolonged to match the timexactivity product of a clinical scan (4 million MBqxs). Spatial resolution was measured on a line source centered on a cylindrical water filled-phantom (18 cm diameter). Count rates of both modalities were compared acquiring a small spherical source at the center of the FOV, in identical time/activity conditions. PET images were generated using the scanner's RAMLA reconstruction, including scatter and attenuation correction; 3 iterations and 8 subsets were used. SPECT images were generated using a custom MLEM and bremsstrahlung-optimized scatter and attenuation correction. 24 iterations were used. Results: Quantitation of the total liver activity shows a 7% error in PET vs a 16% in SPECT. PET spatial resolution is 6.4 vs 15.3 mm for SPECT. The count rate obtained for the spherical source in identical acquisition condition is 6.25 times higher for SPECT than for PET resulting in less noisy images. Conclusions: Both techniques seem to deliver reasonable accuracy. 90Y PET achieves good quantitation results without any bremsstralhung specific correction