Acquisition of Volumetric Models of Skull Base Anatomy Using Endoscopic Endonasal Approaches: 3D Scanning of Deep Corridors via Photogrammetry

Background: Utilization of photogrammetry for creating volumetric models (VM) of neurosurgical dissections and the surgical scene has been limited to image acquisition of surfaces with professional cameras or smartphones. One of the limitations of this 3D scanning technique is the reconstruction of anatomical corridors. We evaluated the feasibility of creating VM of the anterior, middle and posterior skull base and nasal and paranasal fossae anatomy using endoscopic endonasal approaches.

Materials and Methods: Five embalmed and latex-injected human cadaveric heads were dissected through the nasal corridor to expose the anterior, middle and posterior cranial fossa structures and the pterygopalatine fossa. Pictures were taken of different steps of the dissections starting with the sphenoid phase. A rigid endoscope with a 30-degree lens (Stryker, Kalamazoo, Michigan, United States) was used to capture the images from multiple viewpoints via rotation of an arbitrary vertical axis through the center of a structure of interest the relative to the camera at superior, medium, and inferior elevations of the camera, utilized with the “fisheye” lens settings. To improve the texture of the final VM, images were additionally captured with the camera midway from the entrance of the nasal corridor to the skull base structures, with the inclusion of further close-up photos of points of interest. The photographs were taken to achieve that the entire desired surface captured with sufficient partial overlap between the different pictures for alignment. Each surface point is represented in at least two to three images, with one taken in the direct line of the normal to the surface in the region of the location of interest and one to two taken with slight displacement from the normal to the surface in the region of the point of interest. By extension, this necessitates overlap of at least 60% (with ≥ 80% preferred) of a captured area between consecutively taken photographs. With sufficient overlap, the PhotoScan (Agisoft LLC, St. Petersburg, Russia) software can superimpose the acquired images and measure the intersecting tie points to triangulate the location of a point in 3D space, thereby creating a 3D object.

Results: The photogrammetric workflow comprises image capture, preprocessing, postprocessing. An average of 150 photographs was used to construct each model (n = 10). The dense cloud was created from tie points, and the mesh and textures were subsequently added using the PhotoScan. After creating the model, the Blender (Blender Foundation, Amsterdam, the Netherlands) software provided visual improvements, with decimation, smoothing, and texture adjustments of the VM. In the end, we achieved accurate stereoscopic VMs of the nasal corridor and the paranasal fossae and anterior, middle, and posterior fossae structures with a high level of definition and structural integrity including small structures such as nerves and arteries. Moreover, strategic points of interests were labeled and 3D animated for its educational use with a virtual and augmented reality 3D viewer.

Conclusion: The obtained endoscopic VMs represent a new way to depict the anatomy of the skull base, their use with 3D technologies could potentially improve the visuospatial understanding of narrow surgical corridors for education or surgical planning purposes.