Author: Bryan Bednarz, Laura Bennett, Abby E. Besemer, Tyler J Bradshaw, Steve Y Cho, John M. Floberg, Joseph Grudzinski, Elissa Khoudary, Michael J. Lawless 👨🔬
Affiliation: Department of Radiology, University of Wisconsin, University of Wisconsin-Madison Department of Medical Physics, Department of Medical Physics, School of Medicine and Public Health, University of Wisconsin - Madison, Department of Human Oncology, University of Wisconsin School of Medicine and Public Health, Voximetry, Inc, University of Pennsylvania, Department of Radiology, University of Wisconsin - Madison, Department of Human Oncology, University of Wisconsin-Madison 🌍
Purpose: To assess the impact of using a standardized immobilization setup for multi-time-point SPECT/CT imaging on radiopharmaceutical therapy (RPT) dosimetry image registration.
Methods: Ten patients underwent SPECT/CT imaging at ~1, 24, and 96 hrs post-administration of ~200 mCi of Lu-177 vipivotide tetraxetan (Pluvicto®) for the purpose of performing voxelized RPT dosimetry. Five patients were imaged without immobilization and five patients were setup using a standardized immobilization procedure involving a custom head rest, knee triangle, arm and feet straps, and indexing of several anatomic landmarks. The SPECT/CT images at 24 and 96 hrs were registered to the 1 hr reference CT using deformable registration. Registrations were performed in the Torch (Voximetry,Inc) RPT planning software and MIM (MIM Software,Inc). The target registration error (TRE) was calculated for 19 bony and soft tissue landmarks from the head to pelvis in both Torch and MIM. The Hausdorff Distance (HD), Mean Distance to Agreement (MDA), Dice Similarity Coefficient (DSC), and Jaccard Index (JI) were calculated for 25 organs contoured on the reference and deformed CT scans in MIM.
Results: Immobilization reduced average TRE for almost all landmarks for both Torch (p=0.006) and MIM (p=0.1). The HD (p=0.02), MDA (p=0.005), DSC (p=0.01), and JI (p=0.003) contour-based metrics all showed a statistically significant improvement in the accuracy of the normal tissue contour registrations. The registration of the landmarks and contours in the head and neck were most notably improved due to the use of the head rest. Both Torch and MIM performed well when registering most bony anatomy. Challenging areas included the arms/shoulders due to limited field-of-view and the soft tissue in the abdominal cavity, specifically the bowel, due to larger anatomic changes between scans.
Conclusion: A simple standardized immobilization procedure for serial SPECT/CT imaging can significantly improve the image registration accuracy for RPT voxelized dose calculations.