Weighted atlas auto-context with application to multiple organ segmentation

Abstract

Difficulties can arise from the segmentation of threedimensional objects formed by multiple non-rigid parts represented in two-dimensional images. Problems involving parts whose spatial arrangement is subject to weak restrictions, and whose appearance and form change across images, can be particularly challenging. Segmentation methods that take into account spatial context information have addressed these types of problem, which often involve image data of a multi-modal nature. An attractive feature of the auto-context (AC) technique is that a prior "atlas", typically obtained by averaging multiple label maps created by experts, can be used as an initial source of contextual data. However, a prior obtained in this way is likely to hide the inherent multi-modality of the data. We propose a modification of AC in which a probabilistic atlas of part locations is iteratively improved and made available as an additional source of information. We illustrate this technique with the problem of segmenting individual organs in images of pig offal, reporting statistically significant improvements in relation to both conventional AC and a state-of-the-art technique based on conditional random fields.

Multi-part segmentation for porcine offal inspection with auto-context and adaptive atlases

“…When training WAAC models, 32 annotations (10% of a fold’s training pool) were used to compute each weighted atlas ( ). As reported previously, pixel classification accuracy was stable when m w was varied over an order of magnitude [2] .…”

“…In this paper we combine integrated context and WAAC into one system, extending work reported in conference papers on integral context [1] and WAAC [2] . We report a direct comparison of all of these methods applied to segmentation of multiple organs in pig offal, and we also compare with a conditional random field (CRF) method.…”

“…Each training pair consisted of a vector of local and context features and the corresponding class label. When training auto-context methods, performance tended to saturate after five iterations [2] . Therefore we set .…”

Organ Segmentation in Poultry Viscera Using RGB-D

“…More complex approaches with segmentation of multiple organs are less explored. An example is found in [ 17 ], where RGB images of pig offal are segmented into five classes (heart, liver, lungs, diaphragm and an upper portion). This is achieved with a modified auto-context algorithm with an updating atlas, showing a small yet consistent improvement over the regular auto-context algorithm.…”

RGB-D Segmentation of Poultry Entrails

“…Entrails are non-rigid bodies, without straight lines and sharp edges, that only satisfies a weak spatial arrangement. In recent work by [1], a modified auto-context algorithm was developed to segment pig organs in RGB images. The modified algorithm uses an atlas of iteratively updated organ positions.…”