Ball-scale based multi-object recognition in a hierarchical framework
Proceedings of SPIE Medical Imaging 2010: Image Processing, volume 7623, article 762345
Abstract
An abstract is not available in the lab archive. View the paper at its source.
This 2010 SPIE Medical Imaging paper presents a hierarchical method for locating multiple anatomical structures in three-dimensional CT images. It combines prior shape models with intensity-weighted ball-scale information and evaluates recognition on 20 clinical abdominal CT datasets.
Summary
The authors describe a method for recognizing multiple anatomical structures in three-dimensional images without an elaborate search over position, orientation, and scale. It encodes local structure size and image intensity using intensity-weighted ball-scale (b-scale) information, then relates those image patterns to a multi-object shape-model assembly to estimate the assembly’s pose. Evaluation used leave-one-out cross-validation on CT data from 20 patients, focusing on the skin boundary, liver, spleen, and left and right kidneys. The paper reports that recognition depended on the number and distribution of modeled objects; it also notes that the experiments used relatively low-resolution CT images from routine PET/CT scans. Read the paper.
Cite this work
Ulas Bagci, Jayaram K. Udupa, Xinjian Chen (2010). Ball-scale based multi-object recognition in a hierarchical framework. Proceedings of SPIE Medical Imaging 2010: Image Processing, volume 7623, article 762345 https://doi.org/10.1117/12.839920
BibTeX