1706 Objectives: 90Y-radioembolization treatment planning includes determination of the lung shunt fraction (LSF) to maintain mean lung dose to below 25-30 Gy. The estimated mean lung dose constrains the total amount of administered activity and hence the ability to treat tumor to higher dose levels. The device package inserts recommend use of 1000 g lung mass in the calculation of lung dose. Errors in lung mass and lung shunt fraction contributes to the overall errors in lung doses. Here we investigated the accuracy and 95% prediction interval (95%PI) of 5 common methods for calculating the lung mass. The bias can be corrected, if errors are normally distributed, rendering the 95%PI as the more relevant metric. As secondary objective, we investigated the mean errors in lung mass and lung dose that stems from using a fixed 1000 g lung mass for all patients. Relative errors in lung mass estimation are directly proportional to the errors in lung dose. Methods: An IRB-approved retrospective study was conducted on 52 consecutive HCC patients assessed for 90Y-radioembolization with TheraSphere from 2015-2016. As part of routine workup, all patients underwent a diagnostic contrast CT (cCT) of the chest and abdomen prior to the 99mTc-MAA SPECT and non-contrast CT (ncCT) of the liver and chest region. Of these 52 patients, 12 had complete lungs in the SPECT/ncCT field of view (FOV) while 40 had a varying range of lung in the SPECT/ncCT FOV. The gold-standard estimate of true lung mass was defined as the mean of the masses estimated from the cCT and the ncCT using an HU-based volume&density calculation. Linear regression was derived from the complete lung scan data to estimate true lung mass from a lung mass estimated using the cCT HU-based volume&density Methods: This regression was then used on all 52 patients to calculate the true lung mass from the cCT HU-based volume&density mass estimation. The 5 common methods of lung mass calculation investigated were: 1) ncCT HU-based volume&density, 2) ncCT volume & fixed 0.3 g/ml density, 3) fixed mass of 1000 g, 4) gender-based fixed mass of 1000 g for men and 800 g for women, and 5) weight-scaled gender-based fixed mass. A Bland-Altman analysis was performed to estimate the mean relative error and 95%PI of these 5 methods compared to the true lung mass. Results: The mean relative error [95%PI] for lung mass was -9% [±21%] and -11% [±23%] with the ncCT HU-based volumed the errors changed to -23% [±24%] and -14% [±33%], respectively, when partial lung was scanned. In all 52 cases, the fixed mass based methods (1000 g, gender-based, weight-scaled-gender) had large mean errors (26%, 20%, 37%) and very wide 95%PI (±44%, ±41%, ±54%). The errors were not normally distributed in the 1000 g and gender-based fixed mass methods. The 95%PI after bias corrections for all patients were 26%, 31%, and 54% for ncCT HU-based volume&density, ncCT volume&fixed-density, and weight-scaled-gender fixed mass methods. On average, the fixed lung mass of 1000 g overestimated the true lung mass by 23% [±44%] resulting in a corresponding underestimation of the lung dose. Conclusion: Use of the default fixed lung mass of 1000 g severely overestimates the true lung mass. Errors in both ncCT-based methods were lowest compared to fixed mass methods but HU-based volume&density had the narrowest 95%PI. Over all patients, lung mass errors in 95%PI after bias correction with the ncCT HU-based volume&density and volume&fixed-density methods were 26% and 31%. We recommend scanning the maximum lung volume allowed in 99mTc-MAA SPECT/CT and using the bias-corrected ncCT HU-based volume&density method to estimate lung mass when calculating mean lung dose for 90Y-radioembolization treatment planning. Investigations on incorporation of additional errors from LSF estimation on the final mean lung dose is currently underway.