HOW CAN WE IMPROVE THE CURRENT REGISTRATION SYSTEM: PROBLEMS AND SOLUTIONS

Abstract

Navigation is the combination of real and virtual anatomy. Registration brings the virtual world of imaged anatomy into accordance with the real world of actual anatomy. Without a navigation system the process takes place in the head of the surgeon, he assigns the image data to the patient’s anatomy, based on his experience. This process is called mental registration.

Registration methods: Every point in the patient’s anatomy correlates to one point in the three-dimensional image of the patient. Every point in the anatomy and the image can be clearly defined as a position vector in a Cartesian coordinate system. Registration is carried out by a series of transformations of the different Cartesian coordinate systems. The registration between the real and the virtual world can be performed manually or automatically. Different technologies are available for this process.

In the Paired Points Matching landmarks, which can be clearly identified in the image dataset and in the in situ anatomy, are registered pre-operatively in a three-dimensional (e.g. CT) dataset. At least three points, registered as precisely as possible in the dataset and the intra-operative anatomy, are necessary to define the spatial position of the dataset and to bring it into correlation with the patients anatomy. On the spine, the existing prominent landmarks on the accessible dorsal part of the vertebrae, the dorsal process, and the joint condyles are used. Different factors contribute to an inaccurate registration, like an inadequate preparation of the anatomic structure, a misinterpretation of the landmarks by the surgeon, or anatomic variations, that formed in the time between the CT images and the operation.

The definition of corresponding point pairs can be difficult in many applications and an increased degree of invasivity must be accepted. Therefore, the precise recognition of the predefined points of the image dataset in the patient’s anatomy is severely impaired. However, other characteristics, e.g. curves or surfaces of bones can be extracted from the image data. These form the basis for the Surface Matching. A series of points on the surface of the bone must then be digitalised intra-operatively. This accumulation of points is then transferred to the corresponding virtual surface with the help of a complex mathematic algorithm, so that the gap between the points and the surfaces is minimized.

Under special circumstances the registration can be carried out automatically. For this it is necessary that the position of both coordinate systems is known at the time of image recording. To do this, a reference array needs to be attached to the patient and thus the automatic registration can only be performed intra-operatively. In general, all available intra-operative imaging equipment can be used.

Bulky equipment, such as computer tomography or magnetic resonance imaging is available intra-operatively only in very few facilities. The most valuable source for intra-operative images is the image intensifier. Images can be recorded with a navigated, calibrated C-Arm in the standard positions relevant for the surgery. Several fluoroscopic image layers can be displayed at the same time as optical information in the operating room in the form of a permanent virtual fluoroscopy.

Since 2001 a fluoroscopic image intensifier is available, which can generate three-dimensional multilayer reconstructions of high-contrast objects, like bones, from single fluoroscopic images. Since the introduction of three-dimensional imaging techniques in navigation, it is possible to perform the automatic registration of three-dimensional data. So, the above described limitations of CT-based navigation for minimal invasive surgery, e.g. not being able to update the dataset and errors during manual registration, were taken into account. However, the process of automatic registration is highly complex and influenced by many factors.