OBJECTIVE Quantitative imaging of the radionuclide distribution is of increasing interest for microsphere radioembolization (RE) of liver malignancies, to aid treatment planning and dosimetry. For this purpose, holmium-166 ((166)Ho) microspheres have been developed, which can be visualized with a gamma camera. The objective of this work is to develop and evaluate a new reconstruction method for quantitative (166)Ho SPECT, including Monte Carlo-based modeling of photon contributions from the full energy spectrum. METHODS A fast Monte Carlo (MC) simulator was developed for simulation of (166)Ho projection images and incorporated in a statistical reconstruction algorithm (SPECT-fMC). Photon scatter and attenuation for all photons sampled from the full (166)Ho energy spectrum were modeled during reconstruction by Monte Carlo simulations. The energy- and distance-dependent collimator-detector response was modeled using precalculated convolution kernels. Phantom experiments were performed to quantitatively evaluate image contrast, image noise, count errors, and activity recovery coefficients (ARCs) of SPECT-fMC in comparison with those of an energy window-based method for correction of down-scattered high-energy photons (SPECT-DSW) and a previously presented hybrid method that combines MC simulation of photopeak scatter with energy window-based estimation of down-scattered high-energy contributions (SPECT-ppMC+DSW). Additionally, the impact of SPECT-fMC on whole-body recovered activities (A(est)) and estimated radiation absorbed doses was evaluated using clinical SPECT data of six (166)Ho RE patients. RESULTS At the same noise level, SPECT-fMC images showed substantially higher contrast than SPECT-DSW and SPECT-ppMC+DSW in spheres ≥ 17 mm in diameter. The count error was reduced from 29% (SPECT-DSW) and 25% (SPECT-ppMC+DSW) to 12% (SPECT-fMC). ARCs in five spherical volumes of 1.96-106.21 ml were improved from 32%-63% (SPECT-DSW) and 50%-80% (SPECT-ppMC+DSW) to 76%-103% (SPECT-fMC). Furthermore, SPECT-fMC recovered whole-body activities were most accurate (A(est) = 1.06 × A - 5.90 MBq, R(2) = 0.97) and SPECT-fMC tumor absorbed doses were significantly higher than with SPECT-DSW (p = 0.031) and SPECT-ppMC+DSW (p = 0.031). CONCLUSIONS The quantitative accuracy of (166)Ho SPECT is improved by Monte Carlo-based modeling of the image degrading factors. Consequently, the proposed reconstruction method enables accurate estimation of the radiation absorbed dose in clinical practice.