Zu den Inhalten springen

Wir überwinden Grenzen

Geometric Morphometrics and Shape analysis

Contact: Stefan Schlager

The methods of modern morphometrics open a vast field of oportunities for physical anthropologists, as they allow to grasp the shape of a complex configuration of points (representing a biological feature) mathematically. Thus makes it possible to extract statistical information not only adressing a set of points but a shape as a whole. At the moment, we do research in hindsight to medical applications, anthropological questions such as sexual dimorphism, changes of osseous structures due to aging and forensic purposes like reconstructing the shape of soft tissue from information obtained by the underlying bone configuration. To achieve this, I am working with dense sets of coordinates representing surface structures by using sliding semi-landmarks.

Software Development

Most of my current work involves implementing and developing various algorithms into usable software. The platform of my choice hereby is R as it provides flexibility and the integration of C++ and Fortran routines. So far, two packages have been released on my sourceforge page: Morpho and Rvcg. While Morpho mainly targets landmark based shape analysis, Rvcg implements meshing routines from vcglib (Meshlab) into the R environment, thus allowing sophisticated operations from within R. These packages contain the following features:

Morpho

  • Processing landmarks (Im- and Export, Procrustes registration)
  • Various tools, such as detection and removal of outliers based on sample distribution
  • Statistical shape analysis
  • sliding of 3D Semilandmarks on triangular meshes
  • Mesh im/export
  • Mesh warping
  • Visualisation of distances between meshes

Rvcg

  • Mesh smoothing (Taubin smooth, HC Laplacian smooth, standard Laplacian smooth)
  • Mesh simplification (Quadric edge collapse decimation)
  • Removal of isolated pieces(by face count/diameter/only keep largest connected component)
  • Mesh cleaning (removal of duplicated/degenerate faces and vertices)
  • Mesh subsampling (MonteCarlo and Poisson Disk)
  • Fast PLY import
  • Closest point matching
  • Very fast - based on the C++ code from vcglib

Trimesh-tools (additional command line tools)

  • also based on vcglib
  • available from here 

All software is available for Linux (recommended), Mac OSX and Windows.

Examples of our current research

Soft tissue estimation of the human nose

My dissertation thesis tries to extrapolate shape variability of both, soft- and hard-tissue, of the human nose. It is based on surfaces generated from CT-Scans from ~500 Indivuals from Shanghai and Freiburg. The goal is to attribute the shape shifts associated with population membership, gender and age and to incorporate this knowledge into statistical models to estimate the nasal soft-tissue shape from the underlying bone tissue.

Below are some visualisations of the results:

© S. Schlager

Ageing of the soft-tissue nose (green=old; red=young)

© S. Schlager

Population specific asymmetry (exaggerated by factor 10: Green=Chinese;red=European)

© S. Schlager

Skin patch estimated from database (visualised on original skull mesh)

Medical Purposes

© S. Schlager

Construction of titanium mandible plates for reconstruction of the bone structure after tumor section

Deformation based on statistical modelling

Click to play animation © S. Schlager

Extreme pronlongation of a nose based on information from a shape database generated by 250 European noses.The deformation tries to make the prolongation as probable as possible (with probability measured by the distribution of the database’s entries)

Surface registration

Click to play animation © S. Schlager

Two facial surfaces registered by elastic surface matching