In most centres, the choice of the optimal activity to be administered in selective intra-arterial radioembolization with microspheres is nowadays based on empirical models which do not take into account the evaluation of tumour and non tumour individual absorbed dose, despite plenty of published data which showed that local efficacy is correlated to tumour absorbed dose, and that the mean absorbed dose is a toxicity risk factor. A pitfall of the crudest, empirical tumour involvement method are 20 deaths in a single centre which adopted it to administer the whole liver, or the need of systematic 25% subjective reduction of activity prescribed with body surface area method. In order to develop a possibly safer and more effective strategy based on real individual dosimetry, we examine first external beam liver radiation therapy results. The half century experience has something to be borrowed: the volume effect, according to which the smaller the fraction of the irradiated liver volume, the higher the tolerated dose. Different tolerance for different underlying disease or previous non radiation treatment is to be expected. Radiobiological models experience also has to be inherited, but not their dose reference values. Then we report the published dosimetric experience about (90)Y microsphere radioembolization of primary and metastatic liver tumours. In addition we also present original data from our growing preliminary experience of more refined (99m)Tc MAA SPECT based calculations in hepatocarcinoma patients. This overcame the mean dose approach in favour of the evaluation of dose distribution at voxel level. An insight into dosimetry issues at microscopic level (lobule level) is also provided, from which the different radiobiological behaviour between resin and glass spheres can be understood. For tumour treatment, an attenuation corrected (99m)Tc- SPECT based treatment planning strategy can be proposed, although quantitative efficacy thresholds should be differentiated according to the kind of pathology and previous treatment. For non tumour liver parenchyma, data in favour of a relationship between absorbed dose and dangerous effects are encouraging. Unfortunately in hepato-cellular carcinoma, some confounding factors may hamper the adequate estimation of the risk of toxicity. First there is a lack of consensus about the exact definition of toxicity after (90)Y microsphere radioembolization. Second, for HCC patients, progression of both cancer and cirrhosis can simulate a radioinduced toxicity, making the analysis more complex.