One important means of research is volumetric image data as obtained by confocal miscropscopy. By utilizing modern 3D graphics hardware in conjunction with efficient algorithms to process such data, a transition from the traditional cut-based, two-dimensional neuro-anatomy to a real three-dimensional approach becomes possible.

A major goal of our project is to provide a virtual neuro-anatomy laboratory, which integrates all tools in a consistent and convenient framework. This includes tools for data acquisition, data preprocessing, data visualization, image segementation, model/grid generation, data analysis, data transformation and data export. As base system we use the object-oriented, extendable visualization and modelling system Amira which has been developed at ZIB.

The work in this project started with the development of tools to support the generation of standardized 3D brain atlases. Standard brains have been derived for the honeybee and the fruit fly. For this a new method for segmenting confocal microscope scans was used. With this method up to 80% of the time needed for a manual segmentation of the glomeruli in the antennal lobe of the bee, for example, can be saved (cf. Fig. 1). Further on algorithms for automatically extracting thickness data as well as for generating thickness-augmented geometric models have been developed (cf. Fig. 2).

Fig 1: Extrapolation methods allow for fast manual segmentation of anatomical structure. As an example, the images on the left show glomeruli in the antennal lobe. Glomeruli can be identified by comparison between surface reconstruction and the reference model. The image in the middle shows a surface model (colored surfaces) and, superimposed, a volume rendered visualization of the neuropil (brown). The orientation of the model can be changed interactively to match the situation in the experiment. This enables a mapping of activity patterns (right image) to particular structures. Data courtesy: Menzel, Galizia, Brandt.

Fig 2: Reconstructed geometric model of a projection neuron in the bee brain. The projection runs from the antennal lobe to the mushroom body. The model is embedded into the surface reconstruction of some larger anatomical structures (mushroom body). Data courtesy: Menzel

The standard brain of the fruit fly has been used for normalizing gene expression lines from different experiments. Thereby a comparison of the effects of genetic modifications became possible. For co-localization studies, e.g., of gene expression lines in neuropil structure, non-rigid transformations are essential. Good results were obtained with the method described in [4] (cf. Fig. 3). The results that we and the work group of Prof. Menzel attained during practical use are promising. However, problems related to specific properties of confocal microscope data occurred. We pursue approaches that solve these problems. Currently these methods are being implemented and verified. They will be integrated in the Virtual Neuroanatomy Laboratory.

Fig 3: The 3D reconstruction of a Gal4 staining was fitted by means of an elastic deformation to the shape of the mushroom body in the standard brain of the fruit fly. Data courtesy: Heisenberg.

At the present time we develop tools for generating geometric models at the single cell level and augmenting standard brains with functional data, like gene expression lines. We intend to process data from a larger number of concrete biological experiments with these algorithms, thereby verifying them. Numerical simulations of compartment models based on thickness-augmented geometry data have been carried out.

Computer aided neuroanatomy enables quantification and comparison of anatomical findings across work groups. An additional important goal of this project is, therefore, the propagation of methods and tools that have been developed.

• Karlheinz Rein, Malte Zöckler, Michael T. Mader, Cornelia Grübel, and Martin Heisenberg. The Drosophila Standard Brain. Current Biology 2002 12: 227-231.
• Malte Westerhoff (b. Zöckler). Visualization and Reconstruction of 3D geometric models from neuro-biological confocal microscope scans. PhD thesis, FU Berlin, Department of Mathematics and Computer Science, 2003.
• Malte Zöckler, Karl-Heinz Rein, Detlev Stalling, Robert Brandt, Hans-Christian Hege. Creating Virtual Insect Brains with Amira. ZIB-Report ZR-01-32 (2001), http://www.zib.de/PaperWeb/abstracts/ZR-01-32/.
• Malte Zöckler, Detlev Stalling, Hans-Christian Hege. Fast and Intuitive Generation of Geometric Shape Transitions. The Visual Computer, Vol. 16, Issue 5, pp. 241-253, 2000.