liu.seSearch for publications in DiVA
Change search
ReferencesLink to record
Permanent link

Direct link
Analysis of magnetic resonance angiography images using skeletonization and distance transforms
Uppsala University, Sweden.
Linköping University, Department of Medicine and Care, Radiology. Linköping University, Faculty of Health Sciences. Östergötlands Läns Landsting, Centre for Medical Imaging, Department of Radiology UHL.ORCID iD: 0000-0002-7750-1917
2000 (English)In: Discrete Mathematical Problems with Medical Applications / [ed] Ding-Zhu Du, Panos M. Pardalos and Jie Wang, American Mathematical Society (AMS), 2000, 75-89 p.Conference paper (Refereed)
Abstract [en]

Magnetic resonance angiography (MRA) is increasingly performed as a non-invasive method of evaluating patients with suspected vascular disease. In this study we have used images of the arteries of the pelvis obtained after intravenous injection of paramagnetic contrast material, i.e., gadolinium chelates, for arterial enhancement. These images are fairly easy to segment. When interpreting and analysing MRA images, the 3D tree structure and the thickness of the blood vessels are of interest. This shape information may be easier to obtain from the "skeleton" of the blood vessels.

The following image processing steps were performed in the analysis of the blood vessels: resampling the image to cubic voxels, segmentation of the blood vessels from the background through grey-level thresholding and morphological smoothing operations, distance transformation, skeletonization, skeleton pruning (and straightening of zig-zag parts), and finally quantitative (and qualitative) skeleton analysis.

Skeletonization of digital volume objects is either a reduction to a 2D structure consisting of 3D surfaces and curves, or a reduction to a 1D structure consisting of 3D curves only. Thin elongated objects, e.g., blood vessels, are well suited for reduction to curve skeletons. Our skeletonization method first reduces the object to a surface skeleton from which the original object can be recovered. Secondly, the surface skeleton is reduced to a curve skeleton. The topology (i.e., number of components, tunnels, and cavities) and the shape of the object are preserved. The skeletonization is based on a small number of simple local neighbourhood operations, which makes it fairly time efficient. The skeletal voxels are labelled with their (D6) distance to the original background, which in this case conveys information about the local width of the object. Positions for possible artery stenoses may be identified by locating local distance minima in our curve skeletons, which will be investigated further.

Future work will also be directed towards achieving more rotation and noise independent skeletons. We will also develop more "intelligent" and shape preserving pruning methods.

Place, publisher, year, edition, pages
American Mathematical Society (AMS), 2000. 75-89 p.
, DIMACS Series in Discrete Mathematics and Theoretical Computer Science, ISSN 1052-1798 ; 55
National Category
Engineering and Technology
URN: urn:nbn:se:liu:diva-77484ISBN: 0-8218-2096-6ISBN: 9780821820964OAI: diva2:527400
DIMACS Workshop on Discrete Mathematical Problems with Medical Applications, 8-10 December 1999, DIMACS Center, Rutgers University, Piscataway, NJ, USA
Available from: 2012-05-21 Created: 2012-05-21 Last updated: 2014-12-16Bibliographically approved

Open Access in DiVA

No full text

Other links

Search in DiVA

By author/editor
Smedby, Örjan
By organisation
RadiologyFaculty of Health SciencesDepartment of Radiology UHL
Engineering and Technology

Search outside of DiVA

GoogleGoogle Scholar
The number of downloads is the sum of all downloads of full texts. It may include eg previous versions that are now no longer available

Total: 68 hits
ReferencesLink to record
Permanent link

Direct link