Combination Strategies in abdominal Multi-Atlas Image Segmentation

Combination Strategies in abdominal Multi-Atlas Image Segmentation

Combination Strategies in abdominal Multi-Atlas Image Segmentation. The segmentation of organs like the liver, pancreas, and kidneys on abdominal computed tomography (CT) scans can form an input to computer aided diagnosis (CAD) systems and laparoscopic surgery assistance. We display a common, fully automated technique for multi-organ segmentation of abdominal computed tomography (CT) scans. Many existing methods are specialized to the segmentation of individual organs and battle to manage with the variability of the shape and position of abdominal organs. The technique is based on multi-atlas registration, a technique widely used in brain imaging. The obtained probabilistic atlases are used in a graph-cuts based model to obtain a final segmentation.

Image Segmentation

We present amulti-atlas registration framework that is specifically designed to address the challenges found in abdominal scans. To address the high variability in subject-specific appearance as well as the variability in field-of-view. This approach is general to the organ and is able to capture the anatomical variability’s in the available atlas database. Based on a database with manual labels, a target-specific probabilistic atlas is generated for a new subject. To capture inter-subject variability, atlases are refined on three levels on the global level. Based on the defined probabilistic atlas, the final segmentation is obtained by applying a model based on graph-cuts, incorporating high-level.

The principles of atlas is based on  segmentation have been successfully applied. This approach has a major advantage when compared to other segmentation algorithms. Namely, it allows introducing a priori knowledge about the shape and the distribution of the segmented structures in a simple way. The segmentation of organs like the liver, pancreas, computer aided diagnosis (CAD).

Frther this applications include the radiotherapy planning as well as cancer detection and staging. Most of the previous work on automated abdominal segmentation is based on statistical shape models. statistical models are learned on a training set and applied in combination with post processing steps that are often specialized to a particular organ.

Leave a Reply

Your email address will not be published. Required fields are marked *