Rapid Multi-organ Segmentation Using Context Integration and Discriminative Models

Abstract

We propose a novel framework for rapid and accurate segmentation of a cohort of organs. First, it integrates local and global image context through a product rule to simultaneously detect multiple landmarks on the target organs. The global posterior integrates evidence over all volume patches, while the local image context is modeled with a local discriminative classifier. Through non-parametric modeling of the global posterior, it exploits sparsity in the global context for efficient detection. The complete surface of the target organs is then inferred by robust alignment of a shape model to the resulting landmarks and finally deformed using discriminative boundary detectors. Using our approach, we demonstrate efficient detection and accurate segmentation of liver, kidneys, heart, and lungs in challenging low-resolution MR data in less than one second, and of prostate, bladder, rectum, and femoral heads in CT scans, in roughly one to three seconds and in both cases with accuracy fairly close to inter-user variability.

Multi-organ localization with cascaded global-to-local regression and shape prior

“…Submodular strategies may also be designed to reduce the search range (Liu et al, 2010). Lay et al (2013) proposed to increase the organ parameterization complexity by performing a joint anatomical landmarks detection. They obtained accurate and fast results on shapes such as the lungs or the kidneys.…”

Scale-Adaptive Forest Training via an Efficient Feature Sampling Scheme

“…Our method using fine-to-coarse feature sampling is trained at the largest scale δ = 200, which includes all the relevant visual information for the segmentation task. As additional baseline, we perform multi-scale prediction by multiplying the posterior probabilities obtained by the 5 standard forests [10]. Since absolute intensity values are unreliable for MR and ultrasound modalities, we also investigate a variant of the feature space which only allows binarized differences ('Binary' in Table 1, whereas 'All ' denotes the case where the 4 operation types are allowed) to guarantee invariance to changes of illumination and contrast.…”

“…For this reason, incorporating multi-scale information during training is of great interest, and several approaches have been proposed to achieve this objective. A common but computationally costly strategy is to perform several independent learning stages at various scales and combine their outputs at prediction time [6,9,10]. Geremia et al [11] explicitly create a hierarchy of supervoxels and refine the representation when necessary during the forest training.…”

Keypoint Transfer Segmentation

“…Entangled decision forests [13] and a combination of discriminative and generative models [8] have been proposed for the segmentation of CT scans. A combination of local and global context for simultaneous segmentation of multiple organs has been explored [11]. Organ detection based on marginal space learning was proposed in [20].…”