Überatlas: Robust speed-up of feature-based registration and multi-atlas based segmentation
Abstract: Registration is a key component in multi-atlas approaches to medical image segmentation. Current state-of-the-art uses intensity-based registration methods, but such methods tend to be slow and sensitive to large amount of noise and anatomical abnormalities present in medical images. In this master thesis, a novel feature-based registration method is presented and compared to two baseline methods; an intensity-based and a feature-based. The registration method is implemented with the purpose to handle outliers in a robust way and be faster than the two baselines. The algorithm performs a multi-atlas based segmentation by first co-registering the atlases and clustering the feature points in an Uberatlas, and then registering the feature clusters to a target image. The method is evaluated on 20 CT images of the heart and 30 MR images of the brain, with corresponding gold standard. The method produces comparable segmentation results to the two baseline methods and reduces the run-time significantly.
AT THIS PAGE YOU CAN DOWNLOAD THE WHOLE ESSAY. (follow the link to the next page)