Room: Room 205
Purpose: To use contours automatically generated on pre-treatment cone-beam CT (CBCT) to improve the registration with the plan CT for prostate image-guided radiation therapy (IGRT).
Methods: 26 IGRT prostate patients, each with 10 daily kV-CBCTs had their prostate on plan CT and daily CBCT contoured by the same physician (manual contour). Contours are also auto-generated by registering the prior CBCTs to current CBCT deformably and then applying the resulting deformation ï¬?eld to morph the prior manual contours. Then the set of deformed contours on the current CBCT is combined using STAPLE into one optimal set. The average Dice coefficient between auto and manual prostate contours is 86%. IGRT shifts are determined by aligning the centroid of manual contour on the plan CT with the centroid of the manual or auto-contour of the daily CBCTs. These two shifts are then compared to the clinical shift which was based on the therapistâ€™s qualitative alignment of the images. The dosimetric impact of the different shift methods is evaluated by recalculating the dose on each CBCT (assuming a homogeneous density, the plan was first renomalized on the CT also with homogeneous density), and then calculating the coverage of the PTV.
Results: Using manual CBCT contours provided 87.7% PTV coverage, auto-contours 85.7% and clinical only 80.8%. Auto-contours were significantly better then clinical shifts (p=0.0002). Better coverage may be possible using contour overlap registration, rather than just the centroid. CTV (prostate) coverage is >98% for all methods, but PTV coverage is important as this allows for intra-fraction motion, which is common for the prostate. Auto-contours can be produced in under a minute, and the centroid based registration is instantaneous, meaning auto-contour guided IGRT is both more accurate and faster than current standard qualitative IGRT.
Conclusion: IGRT can be fully automated resulting in improved prostate coverage.