Local atlas selection and performance estimation in multi-atlas based segmentation

T.R. Langerak, U.A. Heide, van der, A.N.T.J. Kotte, F.F. Berendsen, J.P.W. Pluim

Research output: Chapter in Book/Report/Conference proceedingConference contributionAcademicpeer-review

5 Citations (Scopus)
1 Downloads (Pure)


In multi-atlas based segmentation, a new image is segmented by registering multiple atlas images and propagating the corresponding atlas segmentations. These propagated segmentations are then combined in a process called label fusion. This paper presents a new, local method that divides the propagated segmentations in multiple, user-definable regions. A label fusion process can then be applied to each of these regions separately and the end result can be constructed out of multiple partial results. The new method is compared to non-local label fusion methods, as well as with another local method called ALMAS. It is shown that local selection does not lead to a significant improvement in cases where existing methods already have a good result, but that our method significantly improves the result of atlas-based segmentation in cases where existing methods are less successful. © 2011 IEEE.
Original languageEnglish
Title of host publicationProceedings of the 8th IEEE International Symposium on Biomedical Imaging : From Nano to Macro (ISBI'11), 30 March - 2 April 2011, Chicago, USA
Place of PublicationPiscataway
PublisherInstitute of Electrical and Electronics Engineers
ISBN (Print)978-1-4244-4128-0
Publication statusPublished - 2011


Dive into the research topics of 'Local atlas selection and performance estimation in multi-atlas based segmentation'. Together they form a unique fingerprint.

Cite this