Very high doses are administered in radiocolloid therapy of cystic craniopharyngiomas. However individual dose planning is not common yet mainly due to insufficient image resolution. Our aim was to investigate whether currently available high-resolution image data can be used for voxel-based dose calculation for short-ranged β-emitters ((32)P,(90)Y,(186)Re) and to assess the achievable accuracy. We developed a convolution algorithm based on voxelized dose activity distributions and dose-spread kernels. Results for targets with 5-40 mm diameter were compared with high-resolution Monte Carlo calculations in spherical phantoms. Voxel size was 0.35 mm. Homogeneous volume and surface activity distributions were used. Dose-volume histograms of targets and shell structures were compared and γ index (dose tolerance 5%, distance to agreement 0.35 mm) was calculated for dose profiles along the principal axes. For volumetric activity distributions 89.3% ± 11.9% of all points passed the γ test (mean γ 0.53 ± 0.16). For surface distributions 33.6% ± 14.8% of all points passed the γ test (mean γ 2.01 ± 0.60). The shift of curves in dose-volume histograms was -1.7 Gy ± 7.6 Gy (-4.4 Gy ± 24.1 Gy for (186)Re) in volumetric distributions and 46.3% ± 32.8% in surface distributions. The results show that individual dose planning for radiocolloid therapy of cystic craniopharyngiomas based on high-resolution voxelized image data is feasible and yields highly accurate results for volumetric activity distributions and reasonable dose estimates for surface distributions.