Session Item

To investigate the contour agreement and dosimetric accuracy in the utilization of different registration methods and anatomical landmarks for cone beam computed tomography (CBCT) verification of primary radical prostate cancer radiotherapy.

15 prostate cancer patients, each with their own 9 sets of CBCT images, were retrospectively recruited. Their planning Computed Tomography (pCT) were deformably registered to their pre-treatment CBCT to construct deformed planning CT (dpCT) images. Therefore, the characteristics of the CBCT images were preserved in the dpCT. All available dpCT images (n=135) were then registered to the pCT using 3D translations and yaw-rotation. The registration was conducted three times for each set of dpCT, using (1) automatic whole image registration (all slices in the range of dpCT dataset), (2) automatic planning target volume (PTV) based registration, and (3) manual registration carried out by an experienced oncologist. Per registration approach, dose distribution in each registered dpCT dataset was calculated with an identical setting to the original treatment plan. Contour agreement between registered dpCT and pCT images was measured for targets and organ at risks (OARs) in terms of dice similarity coefficient (DSC). 

Statistically significant differences in D2% (dose received by 2% volume) and D95% (dose received by 95% volume) PTV, V50Gy (volume in % receiving 50Gy) and V70Gy (volume in % receiving 70Gy) rectum, V55Gy (volume in % receiving 55Gy) and V70Gy bladder, DSC PTV and DSC rectum were found among three registration techniques. The whole image registration significantly reduced the target hotspots (D2% PTV), compared with the PTV-based and manual registrations. Superior PTV and rectal alignments (DSC PTV and DSC rectum) were also observed in the whole image registration than manual registration, and higher performance in bladder preservation (V55Gy and V70Gy) than the PTV-based registration. However, its target coverage (D95% PTV) was inferior to manual registration. The PTV-based registration showed significantly higher D95% PTV, DSC PTV, and DSC rectum with lower rectal dose (V50Gy and V70Gy) than other registration methods. For the manual registration, it had the highest achievement in bladder sparing (V55Gy and V70Gy) among all techniques.

Dosimetric and anatomic improvements can be achieved with a PTV-based registration for prostate cancer patients, increasing target coverage with superior contour agreement and rectal preservation. Target hotspots can be reduced by a whole image registration while manual registration can be employed for a higher level of bladder sparing.