registration. We now produce these masks through image segmentation making the registration completely automatic. We investigate how the accuracy of the segmented masks affects the registration. Materials/Methods: The program was developed using the Insight Toolkit. Prostate patient CTs were segmented via a number of algorithms. By comparing generated to manual contours, Neighborhood Connected was chosen as the preferred segmentation algorithm. Due to the difficulty in segmenting the bladder from prostate, a new mask combining the bladder and prostate volume, was also produced. Three registration methods will therefore be compared: using manual rectum and bladder masks (previous standard), segmented rectum and bladder masks, and segmented rectum and combined bladder prostate masks. Agreement of manual contours of the prostate (which is not masked individually), in the registered images was used to infer registration accuracy. Results: The DC between the segmented and manual masks was up to 96 and 93% for the bladder and rectum respectively. This allowed the segmented mask guided deformable registration algorithm to achieve the same agreement in prostate contours as when manual masks are used, (both average 90% DC). When the combined bladder prostate-segmented mask is used the DC of the prostate contour improves to 95%, (the rectum and bladder contour DC remains the same). Conclusions: A now fully automated hybrid deformable registration method was developed for the prostate region capable of recovering large initial organ variation. Good agreement of segmented masks to the organ they represent is required to achieve acceptable registration results. Using a combined bladder prostate mask is the preferred method as it simplifies the segmentation process, while allowing for improved prostate registration. This automatic registration is used to produce contours and delivered dose for adaptive therapy and cumulative dose for outcome analysis. Author Disclosure: A.R. Godley: None. M. Itomlenskis: None.