TU-C-J-6B-09: Automatic Contour Delineation On Subsequent CT Images Using Deformable Registration

Abstract

Purpose: To implement a deformable registration algorithm to automatically delineate regions of interest (ROIs) on daily CT images by transforming the corresponding ROIs from a reference CT image. Method and Materials: An intensity-based ‘Demons’ deformable registration algorithm was used to find the correspondence between planning CT image and daily CT image. After reference ROIs were delineated manually on the reference CT image, the reference ROIs can be mapped onto each daily CT image. We tested this method on one head-and-neck patient (tonsil tumor). The patient received 3 CT scans per week prior to treatment for a total of 16 CT scans. The deformed ROIs were visually evaluated by a radiation oncologist. The dose-volume histograms (DVHs) of the deformed left and right parotids were calculated and compared with the planned DVHs. In addition, a cumulative dose distribution was calculated by mapping the daily dose distribution from each of the daily CT images back to the planning CT using the deformable registration method. Thus, the DVHs from the cumulative dose distribution can be calculated to represent the final “delivered” dose distribution. The volumetric changes during the elapsed treatment days for targets and critical structures were also investigated. Results: The deformed contours reasonably matched with gross anatomy presented by the CT images. The daily DVHs showed a large variation during the course of treatment due to both setup errors and internal organ deformation. In spite of large variation in DVHs of the parotids in daily CT images, the DVHs of the cumulative dose agreed reasonably well with the planned DVHs for both parotids in this case. Conclusion: We demonstrated that an intensity-based deformable registration algorithm can be used effectively to map the planning ROIs to subsequent CT images acquired during treatment. This allows effortless replanning in an adaptive CT-guided radiotherapy strategy.