Improved Technology for Navigation-Guided Orbital Surgery and Reconstruction

Introduction: There is an increasing interest in endoscopic orbital and transorbital skull base surgery. Navigation guidance is highly important in endoscopic surgery for pathway creation, target manipulation, and reconstruction of normal bone anatomy. A major challenge to orbital navigation is segmentation (partitioning an image into different tissues) of orbital structures such as the globe, optic nerve, and extraocular muscles. The imaging of these structures during navigation is currently suboptimal. In this project, we developed a novel technique of segmenting the orbital bone cavity and registering the bone to the deformed side using intensity based registration. In addition, the optic nerve and extraocular musculature are segmented and transformed using the obtained transformation. These critical structures are included in the surgical planning and intraoperative navigation. In addition, the model can be used to construct patient specific implants or preoperatively.

Method: A software program was created to automatically detect and align contralateral orbital bones to the ipsilateral orbital bones to provide a template for the surgeon to perform accurate reconstruction.

Results: Segmentation: The orbital segmentation was done by progressively locating anatomical landmarks. The initial volume was determined by detecting the nasal tip point and anterior aspect of the frontal bone in tissue and bone threshold volumes, respectively, Fig. 1. The unaffected bone model, was obtained using the detected landmarks the nasal point and finding the zygoma landmark on each slice of the current volume, Fig. 2. Additional anatomical landmarks were detected to segment critical orbital structures.

Registration: The gradient descent based registration was used to rigidly register the unaffected side to the deformed side. In registration, the moving volume was the reflected unaffected side and fixed volume was the deformed side. The gradient descent optimizer was used to minimize the mean square difference between the moving and fixed volumes. Visual inspection and similarity value of transformed volume compared with simply reflecting the volume showed that the registration can achieve more accurate results, Fig. 3. In addition, critical orbital structures segmented from the unaffected side can be transformed to the deformed side with the obtained transformation and be avoided intraoperatively, Fig. 4.

Conclusion: In this paper, we illustrate a method to automatically segment the orbit and orbital structures including the globe, extraocular muscles and optic nerve. This technology is expected to facilitate pre-surgical planning for skull base surgery and allow integration into navigation guidance systems.

Fig. 1 Orbital volume (right) by progressively detecting anatomical landmarks.

Fig. 2 Nasal tip and zygoma landmarks (left) to create the unaffected model (right).

Fig. 3 Simple reflected compare with rigid registration - single representative slice.

Fig. 4 3D point cloud visualization of model registered to the deformed side with transformed critical structures.