Clinical evaluation of atlas-based segmentation for radiotherapy of head and neck tumours

Abstract: Background Semi-automated segmentation using deformable registration of atlases consisting of pre-segmented patient images can facilitate the tedious task of delineating structures and organs in patients subjected to radiotherapy planning. However, a generic atlas based on a single patient may not function well enough due to the anatomical variation between patients. Fusion of segmentation proposals from multiple atlases has the potential to provide a better segmentation due to a more complete representation of the anatomical variation. Purpose The main goal of the study was to investigate potential operator timesaving from editing of atlas-based segmentation compared to manual segmentation for head & neck cancer. Materials and Methods A commercial atlas-based segmentation software (VelocityAI from Nucletron AB) was used together with several expert generated and protocol-based atlases of delineated CT images to create multiple atlas segmentations through deformable registration. The atlas that was considered most universal was selected to construct single atlas segmentation proposals. For fusion of the multiple atlas segmentations an in-house developed algorithm, including information of local registration success was used in a MATLAB-environment1. The algorithm uses weighted distance map calculations where weights represent probabilities of improving the segmentation results. Based on previous results1 the probabilities were estimated using the cross correlation image similarity measure evaluated over a region within a certain distance from the segmentation. Ten patients were incorporated in the study. Each patient was delineated three times, (a) manually by the radiation oncologist, (b) with a single atlas segmentation and (c) with a fusion of multiple atlas segmentations. For the methods (b) and (c) the radiation oncologist corrected the proposed segmentations blindly without using the result from method (a) as reference. For case (c) a total number of 11 atlas segmentations were used. The time spent for segmenting or editing the segmentation proposals by the radiation oncologist was recorded separately for each method and each individual ROI. In addition a grading was used to score how helpful the candidate segmentation proposals were for the structure delineations. The Dice Similarity Coefficient, the Hausdorff distance and the volume were used to evaluate the similarity between the delineated structures. Results The results show a time reduction in the order of 40% when the radiation oncologist only has to correct the multiple atlas-based segmentation proposal compared to manual segmentation. When using single atlas the corresponding figure is 21%. Conclusions Using atlas-based segmentation can reduce the time needed for delineation in the head and neck area of patients admitted for radiotherapy. 1C. Sjöberg and A. Ahnesjö, Evaluation of atlas-based segmentation using probabilistic weighted distance maps, Manuscript, Uppsala University, 2011