TY - JOUR T1 - Automated Internal Dosimetry Research Tool Using Quantitative SPECT for the Lu177 Theranostic Application JF - Journal of Nuclear Medicine JO - J Nucl Med SP - 1301 LP - 1301 VL - 58 IS - supplement 1 AU - Alexander Vija AU - Michal Cachovan Y1 - 2017/05/01 UR - http://jnm.snmjournals.org/content/58/supplement_1/1301.abstract N2 - 1301Objectives: An essential component of optimizing Lu177 therapy is the availability of an automated voxel based dosimetry tool, and minimizing user induced variability. We demonstrate the ability to perform fully automated voxelized dosimetry using quantitative images of Lu177 scans with a specifically developed Dosimetry Research Tool (DRT) integrating dosimetry relevant multi-modal zoning in the reconstruction and dose kernels to compute MIRD dose as reference and a voxelized dose map, as well as an estimated voxelized dose error map using either a mono or bi-exponential fitting model. Subsequent computation of cumulative dose and dose error is enabled by integrating data from all therapy cycles.Methods: We evaluate the correctness of the basic computations using a numerical phantom with known activity and effective decay characteristics for a fit to a mono exponential time activity model using 4 time points. Then, we use NIST traceable calibrated quantitative Lu177 prototype test data of a subject with 4 time points and accompanying CT’s. We automatically segment and label organs, substructures from the reference CT, as well as tumors from the prototype quantitative reconstruction image at the chosen reference time point. This creates a multi modal zone map, which can be used as extra modal information in the reconstruction or post processing. This approach allows for instance to differentiate kidney cortex from medulla using a prototype multi modal reconstruction. Furthermore, we can also include an integrated motion correction and registration scheme. We use dose kernels to compute dose from the quantitative multi modal SPECT reconstructions and extract dose histograms for the VOIs. In addition, we use bootstrapping to estimate the statistical uncertainty in the reconstructions and propagate the errors through the time activity curve fit to compute a dose uncertainty. For comparison we also compute a dose matrix using the MIRD approach.Results: The computation of the voxel based dosimetry of the numerical phantom test object agrees within <2% of truth. The statistical uncertainty estimate in the reconstruction of the phantom lies in the order of 1% per time point and results into dose estimate uncertainty of <4%. The entire computation of a patient data set with 4 time points, including the generation of an automated summary report and dose histogram is within 6 min (excluding reconstruction time).Conclusion: We developed an automated and flexible dosimetry research tool for voxel based dosimetry using quantitative data and CT’s of up to 6 time points. The tool is currently only available to selected research collaborators and is used in an ongoing study to optimize Lu177 protocols. Research Support: ER -