Description
Purpose: To validate and an automated deformable image registration (DIR) algorithm for dose accumulation in organs at risk (OARs) in HDR brachytherapy for cervical cancer. Methods: Image and plan data were collected from 15 patients who underwent 3-5 fractions of HDR brachytherapy to treat cervical cancer. The images were pre-processed to perform a hybrid intensity-based and contour-based DIR: the applicator region and bones were auto-segmented, and together with bladder, rectum/sigmoid and uterus, were automatically masked with a different uniform intensity for each structure. Dice Similarity Coefficient (DSC) and cumulative D2cc of OARs using different fractions as reference were used to assess the accuracy of DIR. The total organ D2cc is averaged considering each fraction as the reference and compared to the total D2cc found by simple DVH addition (clinical method). All DIR steps were done automatically. Results: The average DSC between deformed and reference bladder is 0.93±0.03; and for the rectum the DSC figure is 0.851±0.03. Changes in organ volume as high as 223% still resulted in a DSC of 0.946. On average, the total D2cc using DIR and DVH-based addition differ by 1.918 ± 2.066% for the bladder and 2.257 ± 2.524% for the rectum. However, differences as large as 13.35% in the bladder and 14.46% in the rectum when considering pairs of fractions were observed. The masking and DIR process takes a matter of minutes to complete. Conclusion: We have validated a quick, efficient, and automated DIR algorithm for dose accumulation of multiple fractions for cervical cancer HDR brachytherapy. The algorithm is accurate and simultaneously evaluates all OAR doses. Contour consistency plays an important role in the assessment of the cumulative dose, especially in the rectum. Although the average differences between cumulative D2cc by DIR and simple DVH addition are small, they were found to be significantly high in some patients. In these cases, a more accurate dose calculation is important and can be achieved using our method. More accurate dose calculations lead to the possibility of dose escalation, which could be clinically significant.
