gms | German Medical Science

54. Jahrestagung der Norddeutschen Orthopädenvereinigung e. V.

Norddeutsche Orthopädenvereinigung

16.06. bis 18.06.2005, Hamburg

Statistical shape analysis for spine deformation detection

Meeting Abstract

Search Medline for

  • corresponding author A. Schmitz - Universitätsklinikum Bonn, Orthopädie, Bonn
  • G. Bendels - Bonn
  • M. Smimi - Bonn
  • R. Klein - Bonn

Norddeutsche Orthopädenvereinigung. 54. Jahrestagung der Norddeutschen Orthopädenvereinigung e.V.. Hamburg, 16.-18.06.2005. Düsseldorf, Köln: German Medical Science; 2005. Doc05novEP16

The electronic version of this article is the complete one and can be found online at: http://www.egms.de/en/meetings/nov2005/05nov098.shtml

Published: June 13, 2005

© 2005 Schmitz et al.
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by-nc-nd/3.0/deed.en). You are free: to Share – to copy, distribute and transmit the work, provided the original author and source are credited.


Outline

Text

We describe a medical application where we exploit surface properties (measured in form of 3D-Range scans of the human back) to derive a-priori unknown additional properties of the proband. We perform classification using statistical shape analysis methods.

The data basis consists of 3D-scans taken from 109 adolescent patients undergoing scoliosis treatment. Additionally, in a medical screening co-operation with a local school, we have scanned 165 pupils with no known spine deformation. Before scanning, every proband was examined by an orthopaedist, who also labelled anatomic landmarks with adhesive markers. These anatomic landmarks were chosen for anatomical expressivity and robust detection. The anatomic landmarks themselves form the vertices for a coarse mesh approximation of the back surface recorded in the range scans. In order to capture the geometric variability contained in the back surface, we construct additional landmarks for our mesh, calling Pseudo-Landmarks. After shape alignment the Support Vector Machine classification was applied and the statistical coherence was investigated

of an overall set of single shot scans of the 274 probands.Consistent parameterization and alignment was achieved. Results of the cross validation test revealed a mean precision, linear > 92,1.

We applied statistical analysis in an inter-proband manner, i.e. giving insight over ones

shape characteristics in comparison to the shape space of human backs. The results achieved from the classification algorithm are encouraging that we expect our approach is feasible for screening applications in detection of spine deformities.