Abstract — We present a new method for automatic segmentation of heterogeneous image data that takes a step toward bridging the gap between bottom-up affinity-based segmentation methods and top-down generative model based approaches. The main contribution of the paper is a Bayesian formulation for incorporating soft model assignments into the calculation of affinities, which are conventionally model free. We integrate the resulting model-aware affinities into the multilevel segmentation by weighted aggregation algorithm, and apply the technique to the task of detecting and segmenting brain tumor and edema in multichannel MR volumes. The computationally efficient method runs orders of magnitude faster than current state-ofthe-art techniques giving comparable or improved results. Our quantitative results indicate the benefit of incorporating modelaware affinities into the segmentation process for the difficult case of glioblastoma multiforme brain tumor. Index Terms — Multilevel segmentation, normalized cuts, Bayesian affinity, brain tumor, glioblastoma multiforme.

Abstract. We propose an extended graph-shifts algorithm for image segmentation and labeling. This algorithm performs energy minimization by manipulating a dynamic hierarchical representation of the image. It consists of a set of moves occurring at different levels of the hierarchy where the types of move, and the level of the hierarchy, are chosen automatically so as to maximally decrease the energy. Extended graph-shifts can be applied to a broad range of problems in medical imaging. In this paper, we apply extended graph-shifts to the detection of pathological brain structures: (i) segmentation of brain tumors, and (ii) detection of multiple sclerosis lesions. The energy terms in these tasks are learned from training data by statistical learning algorithms. We demonstrate accurate results, precision and recall in the order of 93%, and also show that the algorithm is computationally efficient, segmenting a full 3D volume in about one minute. 1

Abstract. Locating Brain tumor segmentation within MR (magnetic resonance) images is integral to the treatment of brain cancer. This segmentation task requires classifying each voxel as either tumor or nontumor, based on a description of that voxel. Unfortunately, standard classifiers, such as Logistic Regression (LR) and Support Vector Machines (SVM), typically have limited accuracy as they treat voxels as independent and identically distributed (iid). Approaches based on random fields, which are able to incorporate spatial constraints, have recently been applied to brain tumor segmentation with notable performance improvement over iid classifiers. However, previous random field systems involved computationally intractable formulations, which are typically solved using some approximation. Here, we present pseudo-conditional random fields (PCRFs), which achieve accuracy similar to other random fields variants, but are significantly more efficient. We formulate a PCRF as a regularized discriminative classifier that relaxes the classification decision for each voxel by considering the labels and features of neighboring voxels. 1

In this paper, a semi-automated method for brain hematoma and edema segmentation and volume measurement using computed tomography imaging is presented. The method combines a region growing approach to segment the hematoma and a level set segmentation technique to segment the edema. The main novelty of this method is the strategy applied to define the propagation function required by the level set approach. To evaluate the method, 18 patients with brain hematoma and edema of different size, shape and location were selected. The obtained results demonstrate that the proposed approach provides objective and reproducible segmentations that are similar to the results obtained manually. Moreover, processing time is 4 minutes compared to the 10 minutes required for manual segmentation. Key words: level set segmentation, intracerebral hemorrhage, brain hematoma, brain edema 1.

In this paper, we present a formal theory for symbolically modeling the spatial relationships that exist between gross-level anatomical structures in the human body. We develop these theories with the goal of computer-based inference. The formal theories are used for building models, which can be applied on graph representations of medical images. We describe an end-to-end design for inferring (labeling) parts of the anatomy from an unlabeled three-dimensional data set. Given a finite set of labels (corresponding to anatomical structures from a taxonomy), we probabilistically assign one label to each node in an anatomical graph. The paradigm we present is generalizable and can help bridge the gap between purely informational ontologies and ontologies for intelligent agents. 1

Abstract. This paper proposes a statistical approach to labeling images using a more natural graphical structure than the pixel grid (or some uniform derivation of it such as square patches of pixels). Typically, low-level vision estimations based on graphical models work on the regular pixel lattice (with a known clique structure and neighborhood). We move away from this regular lattice to more meaningful statistics on which the graphical model, specifically the Markov network is defined. We create the irregular graph based on superpixels, which results in significantly fewer nodes and more natural neighborhood relationships between the nodes of the graph. Superpixels are a local, coherent grouping of pixels which preserves most of the structure necessary for segmentation. Their use reduces the complexity of the inferences made from the graphs with little or no loss of accuracy. Belief propagation (BP) is then used to efficiently find a local maximum of the posterior probability for this Markov network. We apply this statistical inference to finding (labeling) documents in a cluttered room (under moderately different lighting conditions). 1

www.elsevier.com/locate/fss