Model-based lower limb segmentation using weighted multiple candidates

First published in
CEUR workshop proceedings 
Number in series
574
Start Page
276
End Page
280
Citation
CEUR Workshop Proceedings 574: 276-280 (2010-12-01)
Contribution to Conference
Workshop Bildverarbeitung für die Medizin, BVM 2010 
Scopus ID
2-s2.0-84893763334
Publisher
Springer
In this paper we propose an extension of active shape model based bone segmentation. We examine the benefit of using multiple candidates for new landmark positions during segmentation. To incorporate this information we compare three strategies of adapting the fitting algorithm. For evaluation we segmented the hip, knee and ankle joints in more than 100 digital radiographs of the lower limbs. We achieve superior accuracy compared to the classic algorithm and prove that segmentation results benefit from multiple candidates.
DDC Class
600: Technik
610: Medizin