Objectives: Voxel-based personalized dosimetry typically requires sequential SPECT images and the co-registered CT images can be used for the attenuation correction (AC), 3D volume-of-interest (VOI) definition and image registration. However, there is potential mis-registration between SPECT and CT images even at the same time point which may hamper the dosimetric accuracy. We aimed to study this effect on AC and VOI definition in this study. Methods: We used the 4D XCAT phantom with In-111 Zevalin activity distribution at 24 hrs post injection, and simulated its 128 noisy SPECT projections over 360° using an analytical MEGP projector, modelling attenuation, scatter, and geometric collimator-detector response. The corresponding attenuation map of the phantom was used as the CT image, which was rotated by -1 to 1 degree in z-axis and shifted by -5 to 5 voxels in the lateral (x-) and anterior-posterior (y-) directions to study the quantitative errors from the misaligned CT image for (i) AC, (ii) VOI definition and (iii) both AC and VOI definition. The noisy projections were reconstructed using OS-EM algorithm of 8 iterations and 16 subsets with full compensation, using differently shifted attenuation maps for AC and mapping out the targeted VOIs, i.e., liver, spleen, kidneys and lungs on the later reconstructed images We calculated the organ activity errors compared to the result using perfectly aligned CT images.

Results: The organ activity errors were about proportional to the degree of shift and rotation. For (i), the errors were larger for organs near ribs (spleen: 3.40% and liver: -2.30% in x-axial shift of 1 voxel). For (ii), the errors were larger for small organs (spleen: -4.57% and kidneys: -4.93% in x-axial shift of 1 voxel) and cold organs (lungs: 3.40% in y-axial shift of -1 voxel). For (iii), the errors were larger for small organs (spleen: -8.32% in x-axial shift of -1 voxel). The z-rotation also impacted small organs more (spleen: -2.56% at -1 degree rotation for (iii)). Conclusions: The mis-registration between SPECT and CT scans would affect both AC and VOI delineation accuracy for organ activities estimation, especially for small and cold organs. These errors would be accumulated for sequential SPECT/CT sessions where there are multiple SPECT and CT mismatches. Thus, image registration should be performed between the corresponding SPECT and CT images for accurate TRT dosimetric calculation.