Share Email Print

Proceedings Paper

Active shape models incorporating isolated landmarks for medical image annotation
Author(s): Tobias Norajitra; Hans-Peter Meinzer; Bram Stieltjes; Klaus H. Maier-Hein
Format Member Price Non-Member Price
PDF $17.00 $21.00

Paper Abstract

Apart from their robustness in anatomic surface segmentation, purely surface based 3D Active Shape Models lack the ability to automatically detect and annotate non-surface key points of interest. However, annotation of anatomic landmarks is desirable, as it yields additional anatomic and functional information. Moreover, landmark detection might help to further improve accuracy during ASM segmentation. We present an extension of surface-based 3D Active Shape Models incorporating isolated non-surface landmarks. Positions of isolated and surface landmarks are modeled conjoint within a point distribution model (PDM). Isolated landmark appearance is described by a set of haar-like features, supporting local landmark detection on the PDM estimates using a kNN-Classi er. Landmark detection was evaluated in a leave-one-out cross validation on a reference dataset comprising 45 CT volumes of the human liver after shape space projection. Depending on the anatomical landmark to be detected, our experiments have shown in about 1/4 up to more than 1/2 of all test cases a signi cant improvement in detection accuracy compared to the position estimates delivered by the PDM. Our results encourage further research with regard to the combination of shape priors and machine learning for landmark detection within the Active Shape Model Framework.

Paper Details

Date Published: 20 March 2014
PDF: 7 pages
Proc. SPIE 9035, Medical Imaging 2014: Computer-Aided Diagnosis, 90353J (20 March 2014); doi: 10.1117/12.2043761
Show Author Affiliations
Tobias Norajitra, Deutsches Krebsforschungszentrum (Germany)
Hans-Peter Meinzer, Deutsches Krebsforschungszentrum (Germany)
Bram Stieltjes, Deutsches Krebsforschungszentrum (Germany)
Klaus H. Maier-Hein, Deutsches Krebsforschungszentrum (Germany)

Published in SPIE Proceedings Vol. 9035:
Medical Imaging 2014: Computer-Aided Diagnosis
Stephen Aylward; Lubomir M. Hadjiiski, Editor(s)

© SPIE. Terms of Use
Back to Top
Sign in to read the full article
Create a free SPIE account to get access to
premium articles and original research
Forgot your username?