Assessment of Dose Accumulation with Different Deformable Image Registration Algorithms for Image-Guided Radiotherapy of Head and Neck Cancer
N Dogan1*, W Sleeman2, D Asher3, M Fatyga4, M Schutzer5, S Song6, (1) University of Miami, Miami, FL, (2) Virginia Commonwealth University, Richmond, VA, (3) Virginia Commonwealth University, Richmond, VA, (4) Mayo Clinic Arizona, Phoenix, AZ, (5) Virginia Commonwealth University, Richmond, VA, (6) Virginia Commonwealth University, Richmond, VASU-E-J-94 Sunday 3:00PM - 6:00PM Room: Exhibit Hall
Purpose:Many automated Deformable Image Registration (DIR) validation methods are based on comparisons of deformed and observer contours. However, contour-based comparisons are not sufficient to assess the dosimetric accuracy of a DIR algorithm. There is also potential for large dosimetric discrepancy when DVH accumulation is performed with different DIR algorithms. In this work, we compare the accumulated doses obtained with two different DIR algorithms in Head and Neck (H&N) cancer.
Methods:Eight H&N patients who had weekly FBCT scans were included. Small Deformation Inverse-Consistent Linear Elastic (SICLE) registration was used was used to estimate the displacement vector fields (DVFs) that simultaneously matched the CT intensity from the source (planning CT) image to weekly FBCT (target) images. In addition to intensity-based matching (SICLE_I), contour-based (SICLE_C) and intensity plus contour-based (SICLE_I+C) matching was also performed. GTV, parotids, larynx and esophagus were used to drive SICLE_C registrations. The SICLE registrations were also compared to ITK Diffeomorphic Demons (ITKDD) registration algorithm. Finally, the accumulated dose was calculated for both adaptive and non-adaptive scenarios and evaluated using dose-volume constraints.
Results:For non-adaptive scenario, there was hardly any difference in the accumulated dose for target coverage using SICLE_I, SICLE_C, SICLE_I+C and ITKDD. Using adaptive replanning, however, PTV1 D98 improved by 1.5% using SICLE_I+C. The maximum accumulated dose to the brainstem (D2) was also reduced by 12% by adaptive. On the average, the left parotid D50 also decreased by 32% using adaptive replanning. The maximum dose to the cord did not have a significant change using adaptive replanning-based dose accumulation using all DIR algorithms.
Conclusion:Intensity driven DIR algorithms used in this work produced fairly similar accumulated dose results in H&N cases. Overall, the dose accumulation based on adaptive replanning and utilizing intensity + contour driven registration resulted in better target coverage and critical structure sparing.