A Low-Rank Sparse Approach to Reconstruct Incomplete Shapes

Statistical Shape Models are often used as shape priors in medical image analysis. For the construction of reliable SSMs it is usually required to have a dataset of images which is big enough to cover all the possible shape variations in a population. To this end, we have developed a method that allows to build SSM’s from already available databases: routine clinical data [2]. Challenges of clinical data are among others acquisition artifacts or incomplete shapes. Our initial application was building SSMs of skulls from CT scans. The proposed method makes use of low-rank and sparse matrix decomposition techniques to automatically select only the uncorrupted parts of each shape in the construction of the SSM. Here we present a variation of this algorithm adapted to the task of reconstructing liver shapes. Our results show that our method is able to reconstruct incomplete shapes with an average mean distance of 2.3mm.     «

Statistical Shape Models are often used as shape priors in medical image analysis. For the construction of reliable SSMs it is usually required to have a dataset of images which is big enough to cover all the possible shape variations in a population. To this end, we have developed a method that allows to build SSM’s from already available databases: routine clinical data [2]. Challenges of clinical data are among others acquisition artifacts or incomplete shapes. Our initial application was b...     »