Segmenting Brain Tumors Using Pseudo–Conditional Random Fields

Abstract

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.

3D blob based brain tumor detection and segmentation in MR images

“…Therefore, blobs that have a bilateral match can be ) applied to C. The three thresholds that achieve the highest recall rates (γ 4,5,6 ) are labeled in the plot and their detailed outcomes are shown in Table 1. discarded as they are likely to be normal blob-like structures of the brain.…”

“…The majority of the recent work in this area focuses on supervised learning methods for single tumor segmentation [3][4] [5][6] [7]. Supervised methods require excessive computation time for training on fully registered 3D brain scans, and a detailed comparison of prior work shows that automatic tumor segmentation can take up to hours per 3D scan [3].…”

Iterative Probabilistic Voxel Labeling: Automated Segmentation for Analysis of The Cancer Imaging Archive Glioblastoma Images

“…To address these deficiencies, effort has been devoted to developing automated algorithms for segmenting tumor volumes. [8][9][10][11][12] These algorithms include clustering, 13,14 discriminative strategies, 15 and generative approaches. 11,16,17 The success of these methods has been limited by widely differing MR imaging protocols for image acquisition and quality 18 and the significant overlap between the radiographic appearance of glioblastoma tumors and normal cerebrum on MR imaging.…”

Joint Tumor Segmentation and Dense Deformable Registration of Brain MR Images

“…State of the art segmentation methods combine efficient classification techniques [3] with low level segmentation methods [4]. From such perspective, tumor detection is addressed as a classification problem where one aims at separating healthy from diseased tissues at the voxel level, while imposing smoothness constraints.…”