Concurrent tumor segmentation and registration with uncertainty-based sparse non-uniform graphs

Abstract

In this paper, we present a graph-based concurrent brain tumor segmentation and atlas to diseased patient registration framework. Both segmentation and registration problems are modeled using a unified pairwise discrete Markov Random Field model on a sparse grid superimposed to the image domain. Segmentation is addressed based on pattern classification techniques, while registration is performed by maximizing the similarity between volumes and is modular with respect to the matching criterion. The two problems are coupled by relaxing the registration term in the tumor area, corresponding to areas of high classification score and high dissimilarity between volumes. In order to overcome the main shortcomings of discrete approaches regarding appropriate sampling of the solution space as well as important memory requirements, content driven samplings of the discrete displacement set and the sparse grid are considered, based on the local segmentation and registration uncertainties recovered by the min marginal energies. State of the art results on a substantial low-grade glioma database demonstrate the potential of our method, while our proposed approach shows maintained performance and strongly reduced complexity of the model.

Comparative Evaluation of Registration Algorithms in Different Brain Databases With Varying Difficulty: Results and Insights

“…4 for example). An ideal registration approach should accurately align the normal regions (which do have correspondences across images), and relax the deformation in the pathology-affected regions, where no correspondences can be found [26], [27]. Literature has suggested to either mask out the pathological regions from the registration process (i.e., the cost-function-masking approach [27]), or, to simulate a pathological region in the normal-appearing template (i.e., the pathology-seeding approach [28]–[34]).…”

(Hyper)-graphical models in biomedical image analysis

“…The same principle was used in [44] to deal with the task of image-to-volume registration through a decoupled inter-connected graphical model seeking to simultaneously determine the optimal plane through a higher order model and the associate in plane deformation through a pair-wise graph as suggested in [32]. The problem of coupled segmentation/atlas-based registration within the context of brain tumor delineation was considered within an adaptive graphical model in [45]. The central idea was to consider hypercliques modeling the interacting segmentation and deformation labels and adaptively resample the associating grids while explicitly accounting for the uncertainties, as suggested in [32], of the obtained solution.…”

Modeling 4D pathological changes by leveraging normative models

“…A fourth class of methods is joint segmentation and registration (Pohl et al, 2006; Parisot et al, 2014; Kwon et al, 2014; Liu et al, 2014). Pohl et al (2006) proposed a Bayesian framework for joint segmentation and registration for normal brain MR images.…”