Hierarchical Hybrid Shape Representation for Medical Shapes

Abhishek Kolagunda, Guoyu Lu and Chandra Kambhamettu

Abstract

Advances in 3D medical imaging technology have led to increase of interest in shape analysis of organs. This has in turn led to explosion of 3D medical shape data collected. The 3D shape data is also being used for simulations and to guide minimally invasive and remote surgical procedures. We present an Hierarchical-Hybrid Shape Representation (HSSR) which is compact and has both explicit and implicit forms. The compactness of the representation largely reduces storage requirement and communication overheads. The explicit and implicit forms can be used for accurate visualization of organ shapes and guide surgical procedures. The hybrid shape model proposed is a combination of Extended Superquadrics and RBF interpolation function that models, separately, the base shape and surface deformations. We also present an automatic method to fit the hybrid shape model to complex shapes by hierarchically dividing the shape into parts. Finally, we propose a technique to reconstruct shape from its compact representation by recursively blending the parts using intersection shape. Our extensive experiments show that our shape representation method significantly outperforms existing approaches in both accuracy and compactness.

Session

Poster 1

Files

Extended Abstract (PDF, 446K)
Paper (PDF, 1201K)

DOI

10.5244/C.29.74
https://dx.doi.org/10.5244/C.29.74

Citation

Bibtex

@inproceedings{BMVC2015_74,
	title={Hierarchical Hybrid Shape Representation for Medical Shapes},
	author={Abhishek Kolagunda and Guoyu Lu and Chandra Kambhamettu},
	year={2015},
	month={September},
	pages={74.1-74.13},
	articleno={74},
	numpages={13},
	booktitle={Proceedings of the British Machine Vision Conference (BMVC)},
	publisher={BMVA Press},
	editor={Xianghua Xie, Mark W. Jones, and Gary K. L. Tam},
	doi={10.5244/C.29.74},
	isbn={1-901725-53-7},
	url={https://dx.doi.org/10.5244/C.29.74}
}