Improving Probabilistic Image Registration via Reinforcement Learning and Uncertainty Evaluation
One framework for probabilistic image registration involves assigning probability distributions over spatial transformations (e.g. distributions over displacement vectors at each voxel). In this paper, we propose an uncertainty measure for these distributions that examines the actual spatial displacements, thus departing from the classical Shannon entropy-based measures, which examine only the probabilities of these distributions. We show that by incorporating the proposed uncertainty measure, along with features extracted from the input images and intermediate displacement fields, we are able to more accurately predict the pointwise registration errors of an intermediate solution as estimated for a previously unseen input image pair. We utilize the predicted errors to identify regions in the image that are trustworthy and through which we refine the tentative registration solution. Results show that our proposed framework, which incorporates uncertainty estimation and registration error prediction, can improve accuracy of 3D image registrations by about 25%.
KeywordsUncertainty Measure Image Registration Registration Error Target Registration Error Deformable Image Registration
Unable to display preview. Download preview PDF.
- 4.Grady, L.: Multilabel random walker image segmentation using prior models. In: IEEE CVPR, pp. 763–770 (2005)Google Scholar
- 10.Lotfi, T.: Uncertainty in probabilistic image registration. Master’s thesis, Simon Fraser University, Burnaby, Canada (2013)Google Scholar
- 11.Muenzing, S., Murphy, K., Ginneken, B., Pluim, J.: Automatic detection of registration errors for quality assessment in medical image registration. In: SPIE Medical Imaging: Image Processing, vol. 7259, pp. 1–9Google Scholar
- 12.Muenzing, S., van Ginneken, B., Pluim, J.P.W.: DIRBoost: An algorithm for boosting deformable image registration. In: Biomedical Imaging (ISBI), pp. 1339–1342 (2012)Google Scholar