Purpose: To evaluate the effect of image registration type and quality in post-radioembolization Y90 PET/CT dosimetry
Methods: Nine consecutive liver-transplant patients with 15 liver tumors who received Y90 radioembolization were included for Y90 PET/CT based post-implant dosimetry analysis. All lesions had pathological response data available. The tumors and liver were contoured by a radiation oncologist on diagnostic MRI or CT images that were fused with post-implant Y90 PET/CT images to evaluate the delivered dose distribution using a commercial algorithm (MIM Software Inc., Cleveland, OH). Contours were propagated using either rigid registration focused on tumor alignment or deformable registration with and without local vector refinements. Dose volume histogram (DVH) data for tumors and normal liver were analyzed to evaluate the dosimetric effects of fusion quality. Standard metrics to quantify registration quality (dice indices, landmark registration errors) are also reported.
Results: Large variations in absolute differences were observed between the rigid and deformable registration based doses to the tumor [∆D90 = 78.1% (±153.9%), ∆V150Gy = 8.2% (±11.3%), mean tumor dose difference = 20.3% (±13.8%)] and liver [mean-dose difference = 10.9% (±18.2%)]. No dosimetric differences were observed between the default global and locally refined deformable registrations. Significant enhancements in registration quality were noted [mean liver dice indices improved by > 12% (±7%), mean registration error reduced by > 5 (±1) mm] for global deformable registrations relative to rigid registrations. Improvements in deformable fusion quality with local refinements were modest and had negligible dosimetric impact. Deformable registration improved the correlation between dosimetric parameters and pathological response.
Conclusion: Improvements in fusion quality with deformable registration can lead to large differences in tumor and liver Y90 post-implant dosimetry compared to rigid registration. In our study, the use of local refinements to improve deformable registration quality did not significantly influence the post-implant dosimetry.
Image Fusion, PET, Dose Response