Room: Karl Dean Ballroom C
Purpose: We present a QA metric for MVCT-to-kVCT deformable image registration (DIR) which compares the region-of-interest (ROI) DVHs from the two images. To assess effectiveness, we compare this technique with Jacobian determinant (|J|) maps calculated from the same image pair.
Methods: The plan kVCT and daily IGRT MVCT images were obtained from prostate, head-and-neck, and abdomen patients. The dose distribution was calculated upon each daily MVCT set using the planned beam intensity pattern. DIR was applied between the MVCT and plan kVCT; the plan ROIs were warped onto the MVCT, and the MVCT dose distribution was warped onto the plan kVCT. For selected ROIs, DVHs from the MVCT and kVCT data sets were overlaid. For a “perfect� DIR, both DVHs should align perfectly; DVH differences may indicate DIR complications within the ROI. The ROIs were superimposed upon |J| maps compiled for the DIR, for comparison.
Results: For a prostate case, good agreement between the DVHs for all ROIs was observed; the |J| map revealed the mean values within the ROIs to be approximately unity. For a head-and-neck case, the DVHs were nearly identical for all ROIs considered except the esophagus; the |J| value was slightly elevated within the esophagus between the lungs at the anterior aspect. For an abdomen case with interfraction weight loss, the DVHs were nearly identical for all ROIs considered except stomach and liver. For those ROIs, |J| values exceeding unity were observed; within the stomach at the patient-right aspect, |J| > 2 was observed, indicating that the DIR may have “struggled� in those regions.
Conclusion: DVH overlays can effectively determine DIR spatial accuracy. For ROIs with severe deformation as indicated by |J| maps, DVH overlays can identify residual DIR inaccuracy. An advantage of DVH overlays is their more direct relevance when assessing ROI coverage or avoidance.