(Sunday, 7/29/2018) 3:00 PM - 6:00 PM
Room: Exhibit Hall
Purpose: Quantifying therapeutic induced intra-tumoral FDG uptake changes at the image voxel level requires deformable image registration (DIR) performed between pairwise PET/CT images. This study aims to investigate the DIR algorithms’ performance on registering sequential PET/CT images obtained from a human head-and-neck cancer xenografts tumor undergoing radiotherapy.
Methods: Xenografts were established in 15 nude NIH III mice by injecting UT-SCC-14 cells subcutaneously into the flank. The mice were irradiated by delivery with 2Gy/day and 5days/week. 18F-FDG micro-PET/CT scans were performed before treatment and after 10Gy and 20Gy of dose delivery with a reconstructed PET/CT image voxel size 0.5x0.5x0.6/0.15x0.15x0.15mm³. Tumor was manually delineated on PET/CT scans. Weekly PET/CT images were aligned manually to their pre-treatment PET/CT images based on the tumor contours. An image intensity based DIR (IM-DIR) and a biomechanical based DIR (BM-DIR) were performed respectively between each weekly CT and its pre-treatment CT images using the tumor contours as constraints or boundary conditions. The transformations were applied to deform weekly PET images to their pre-treatment images. The Jacobian and the standard uptake value (SUV) obtained using the two DIRs were compared. To test the effect of tumor volume changes on the registration discrepancy, Spearman’s correlation was performed between the tumor volume changes and the average SUV absolute differences from the two DIRs.
Results: Tumor volume increased by (39.84±17.2)% after 20Gy of dose delivery. The Jacobian values were 1.28±0.73 and 1.28±0.47 for the IM-DIR and BM-DIR. The displacement-vector-difference between the two DIRs was (0.39±0.33)mm and the SUV absolute difference was (7.4±9.02)%. The correlation between the tumor volume changes and the tumor voxel SUV discrepancies was not significant (p=0.107).
Conclusion: The BM-DIR produced more physical plausible transformations with less voxels have extreme Jacobian values. The tumor voxel SUV discrepancies caused by the two DIRs were uncorrelated to the tumor volume changes.
Keywords
FDG PET, Finite Element Analysis, Functional Imaging
Taxonomy
IM/TH- Image registration : PET
Contact Email
