Shape Analysis and Interactive Shape Space Exploration

Abstract

Details

Description

The project focuses on morphological analysis and visualization methods for comparative morphology of biological structures. In this context the data is provided as micro-CT scans, as they provide high-resolution images, are non-destructive and can thus also be used for valuable material.

The projects tasks are structured as follows:

  • Methods for mean shape computation. This part comprises automatic techniques for 3D registration and dense matching between the individual datasets. New methods are needed to incorporate varying characteristics within the datasets, e.g. differently worn-out teeth within the same class of rodents should not influence the mean shape. Since such aspects are highly specific to our application the shape metrics has to be tuned to our particular cases. To this extend user-assisted as well as automatic techniques have to be explored.
  • Methods for visual exploration of a shape space. To facilitate morphological studies it is envisaged to develop tools that enable an interactive, unrestricted visual navigation of the high dimensional manifold spanned by the collected shape samples (similar to techniques developed in the context of image databases). Since shapes possess a meaningful and intuitive native presentation form, which is in stark contrast to other cases of high dimensional exploration, a navigational approach can serve for the detection of correlations and dependencies as well as for forming of new hypotheses.
  • Visualization techniques for conveying shape variance. Techniques need to be investigated and developed which visualize the very complex morphology and variation of rodent skulls. A tight coupling of these techniques with the visual exploration tool is intended e.g. for selection of a subset of specimen in the exploration tool for which the variance is then visualized. Selection of subsets of specimens along with close-ups of the selected subset in the exploration tool allow more detailed exploration of certain morphological features and the identification of the specimens, populations, or species involved in particular subspaces.

Of course, the developed methods are not restricted to the specific application area but can be applied to all other fields of research where shape analysis and visualization is required.

Since 2012 this project is part the externDFG SPP 1335 on "Scalable Visual Analytics".

From 2009 to 2012 this work was funded by a externB-IT Research School scholarship.

Publications

 
Max Hermann, Anja C. Schunke, Thomas Schultz, and Reinhard Klein
In: IEEE Trans. on Visualization and Computer Graphics (2016), 22:1(708-717)
 
In: Computers & Graphics (Sept. 2015), 53, Part A(63-71)
 
Max Hermann, Anja C. Schunke, and Reinhard Klein
Poster at Symposium on Statistical Shape Models & Applications (SHAPE2014) in Delémont, Switzerland, June 2014
 
Max Hermann, Anja C. Schunke, Thomas Schultz, and Reinhard Klein
In proceedings of IEEE PacificVis 2014, Mar. 2014
 
In proceedings of Central European Seminar on Computer Graphics for Students (CESCG'2013), pages 113-120, Apr. 2013
 
In proceedings of 2012 ACM SIGGRAPH/Eurographics Symposium on Computer Animation, Lausanne, Switzerland, July 2012
 
Max Hermann, Anja C. Schunke, and Reinhard Klein
In proceedings of BioVis 2011: 1st IEEE Symposium on biological data visualization, Providence, Rhode Island, USA, pages 151-158, IEEE, Oct. 2011
 
Anja C. Schunke, Max Hermann, and Reinhard Klein
Poster presentation at 9th Int. Congress of Vertebrate Morphology in Punta del Este, Uruguay, July 2010

Projects and Software

  • StarVis. StarVis is an application for explorative investigation of similarities in morphometric attributes between different geographical populations. Based on glyph visualization of morphometric attributes overlayed on a geographic map it features dynamic views and interactive filtering. Similar groups can be visually clustered on several scales using Voronoi diagrams.
    See StarVis project page details and software.

  • Semantically Steered Visual Analysis of Shape Space. Relying on densely registered volumetric shape datasets (such as acquired in modern micro CT) we developed interactive methods which, exploiting semantic knowledge, allow a meaningful investigation of shape space. The techniques are efficient and targeted at highly detailed morphometric shapes. First promising results are achieved in analysis of rodent mandibles which provide a challenging example regarding their high shape variability.
    See project page for details.
Screenshot of StarVis application.
Overview of visual analysis system for morphometric shape spaces.