The bony eye socket (orbita) embedding the eye ball (bulbus) as well as the ocular muscles has the shape of a funnel or more precisely that of a four sided pyramid. The orbital content is bounded and thus protected by the orbital floor, the roof, and the lateral walls (cf. Fig. 1). The frontal opening is representing the pyramidal base and the dorsal opening at the tip is a duct for the optical nerve connecting the retina with the brain. The orbital rim represents a stable frame conducting forces around the orbit, while its walls, to some extent only a few tenth of a millimeter in thickness, are thin as paper (cf. Fig. 2).

Orbital malformations can have different reasons like congenital dysplasia, fractures due to trauma (Fig. 3), or defects caused by inflammation or tumor resection. The surgical task is to re-establish the orbital function, i.e. to reconstruct the orbital walls under consideration of the orbital volume, and to achieve a symmetric and physiological correct position of the eyes. In addition to the functional rehabilitation an esthetic rehabilitation is strived as well. Thus, the therapeutic concept has to be planned thoroughly.

With spiral CT multislice scanners and ever increasing spatial resolution CT imaging became the method of choice for fracture diagnostics. Fractures of the orbital floor, for instance, are clearly visible in coronal or sagittal CT slices (cf. Fig. 3 left). Transversal slices are better suited for the assessment of lateral wall defects and asymmetries. However, for the surgical reconstruction of a malformed orbit three-dimensional visualizations of the anatomy are required. 3D models of individual anatomy allow for a planning of bone relocations or the customization of implants like cover sheets or micro-meshes.

For a 3D reconstruction of individual anatomy from CT data computer-assisted segmentation tools are needed. Threshold based segmentation of the orbits with its thin bone structures lead to insufficient reconstruction results due to partial volume effects (cf. Fig. 4). Thus, more sophisticated segmentation tools as provided, for instance, with amira are needed.

With manual segmentation using intelligent tools we are able to reconstruct adequate orbit models from CT data (cf. Fig. 5). However, since the segmentation is tedious, the aim of the Orbita project is to develop segmentation tools that support that segmentation process. The idea is to gather knowledge of normally developed orbits and transfer it into a statistical 3D shape model. This model, a so called atlas, contains a mean shape of the orbit's anatomy including all of its variations within a normal range. Our long-term objective is to automatically reconstruct thin bony structures like the orbital walls with the help of anatomic knowledge being represented in such atlases of the respective anatomy.

In a first study we try to establish such a normalized model of the human orbit. Therefore, we are going to segment a selected set of CT data sets containing non-pathological orbits. Key anatomical landmarks as well as a standardized topology are to be defined. Afterwards the orbital surface models are similarly parameterized for a 3D shape analysis. Having such a shape model including its relevant geometric variations we are able to fit it as close as possible to iso-surfaces or image gradients, steering the segmentation process, or even use it as a reconstruction template for the design of implants or transplants as it is the goal for other parts of the skull, as for instance the midface, mandible or the neuro-cranium .

• Hans Lamecker, Lukas Kamer, Antonia Wittmers, Stefan Zachow, Thomas Kaup, Alexander Schramm, Hansrudi Noser, Beat Hammer. A method for the three-dimensional statistical shape analysis of the bony orbit. Proc. Computer Aided Surgery Around the Head, p. 94–97, 2007.
• Lukas Kamer, Hansrudi Noser, Hans Lamecker, Stefan Zachow, Antonia Wittmers, Thomas Kaup, Alexander Schramm, Beat Hammer. Three-dimensional statistical shape analysis - a useful tool for developing a new type of orbital implant?. AO Foundation, 2006.