Compliance with radiobiologic principles of radionuclide internal dosimetry is fundamental to the success of 90 Y radioembolization. The artery-specific SPECT/CT partition model is an image-guided personalized predictive dosimetric technique developed by our institution, integrating catheter-directed CT hepatic angiography (CTHA), 99m Tc-macroaggregated albumin SPECT/CT, and partition modeling for unified dosimetry. Catheter-directed CTHA accurately delineates planning target volumes. SPECT/CT tomographically evaluates 99m Tc-macroaggregated albumin hepatic biodistribution. The partition model is validated for 90 Y resin microspheres based on MIRD macrodosimetry. Methods: This was a retrospective analysis of ourearly clinical outcomes for inoperable hepatocellular carcinoma. Mapping hepatic angiography was performed according to standard technique with the addition of catheter-directed CTHA. 99m Tc-MAA planar scintigraphy was used for liver-tolung shunt estimation, and SPECT/CT was used for liver dosimetry. Artery-specific SPECT/CT partition modeling was planned by experienced nuclear medicine physicians. Results: From January to May 2011, 20 arterial territories were treated in 10 hepatocellular carcinoma patients. Median follow-up was 21 wk (95% confidence interval [CI], 12-50 wk). When analyzed strictly as brachytherapy, 90 Y radioembolization planned by predictive dosimetry achieved index tumor regression in 8 of 8 patients, with a median size decrease of 58% (95% CI, 40%-72%). Tumor thrombosis regressed or remained stable in 3 of 4 patients with baseline involvement. The best a-fetoprotein reduction ranged from 32% to 95%. Clinical success was achieved in 7 of 8 patients, including 2 by sublesional dosimetry, in 1 of whom there was radioembolization lobectomy intent. Median predicted mean radiation absorbed doses were 106 Gy (95% CI, 105-146 Gy) to tumor, 27 Gy (95% CI, 22-33 Gy) to nontumorous liver, and 2 Gy (95% CI, 1.3-7.3 Gy) to lungs. Across all patients, tumor, nontumorous liver, and lungs received predicted $91 Gy, #51 Gy, and #16 Gy, respectively, via at least 1 target arterial territory. No patients developed significant toxicities within 3 mo after radioembolization. The median time to best imaging response was 76 d (95% CI, 55-114 d). Median time to progression and overall survival were not reached. SPECT/CT-derived mean tumor-to-normal liver ratios varied widely across all planning target volumes (median, 5.4; 95% CI, 4.1-6.7), even within the same patient. Conclusion: Imageguided personalized predictive dosimetry by artery-specific SPECT/CT partition modeling achieves high clinical success rates for safe and effective 90 Y radioembolization. Asaformofart erial territory-specific point-source brachytherapy, 90 Y radioembolization is always effective when delivered at the right location, in the right dose, and with the right intent. 90 Y radioembolization failure is invariably due to one or a combination of these 3 factors being incorrectly addressed. Responsibility for this triad of factors is ...