[en] Full text of publication follows. Aim: our purpose was to compare different Methods of calculating ⁹⁰Y microspheres activity for liver treatment by Internal Selective Radiation Therapy (SIRT). Such comparison is not yet available and is needed in clinics to optimize patient specific dosimetry. We also investigated lungs breakthrough (LB) calculation and its impact on the maximum injectable activity. Materials and Methods: 31 consecutive evaluations based on ^99mTc macro-aggregates (MAA), followed by 25 treatments with ⁹⁰Y microspheres SIR-spheres^R (SIRTEX) were performed. Tumor and healthy liver volumes were determined by the interventional radiologist on anatomical images acquired with a standard contrast enhanced CT (ceCT) protocol. To determine the Tumor to Normal liver uptake ratio (T/N), regions of interest (ROIs) were drawn on ^99mTc MAA SPECT/CT images acquired with a hybrid gamma camera Infinia-Hawkeye (GE). The ROIs were drawn by both a nuclear physicist and a physician via a triple fusion between SPECT/CT data, ceCT and either PET or MRI. For each treatment, four different Methods of calculating ⁹⁰Y activity were applied retrospectively: 3 based on Body Surface Area (BSA-1, BSA-2 and BSA Kennedy) and one based on MIRD formalism (Partition Model). Relationships between calculated activities, LB, T/N ratio and tumor involvement were investigated. In the same way, lobar and total liver treatments were analysed separately. Results: when attenuation correction was not considered, overestimation of LB was on average 65%, but in any case the estimated lungs' doses were below 30 Gy. Moreover, LB was not significantly related to the T/N ratio, neither to tumor involvement nor Radiochemical Purity. Differences in calculated ⁹⁰Y activities were extremely large, being greater for lobar treatments (from -85% to 417%) than whole liver treatments (from -49% to 58%). Besides, 2 values of T/N ratio were identified as thresholds: one for BSA-based Methods (average dose to healthy liver is higher than 30 Gy for T/N ≤ 3) and one for partition model (average dose to tumor is higher than 120 Gy for T/N ≥ 4). Conclusion: the maximum injectable activity is not limited by lung irradiation. As the partition model accounts for differences in uptake between normal and tumour liver, thus optimizing the risk-benefit ratio, this method should be preferred over BSA-based Methods. Nevertheless, partition model assumes an adequate match between MAA and ⁹⁰Y microspheres that should be investigated with further studies. Calculating doses with partition model partially introduces a bias that would be removed using a gold standard method such as Monte Carlo simulations. (authors)