Chapter 10 - A Context Integration Framework for Rapid Multiple Organ Parsing

Abstract

We propose a context integration 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 overall volume patches, while the local image context is modeled with a local discriminative classifier. Through nonparametric modeling of the global posterior based on approximate nearest neighbors, it exploits sparsity in the global context for efficient detection. The approximate nearest neighbors are found by constructing a forest of random binary space partition trees and using the Region Clustering criteria that encourage contiguity and locality. 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. Further, the framework is theoretically extended to incorporate object context for exploiting contextual information at object level. Using our approach, we demonstrate efficient detection and accurate segmentation of liver, kidneys, heart, and lungs in challenging low-resolution magnetic resonance data in less than 1s, and of prostate, bladder, rectum, and femoral heads in computed tomography scans, in roughly 1-33s and in both cases with accuracy fairly close to inter-user variability.