METHOD OF SEED ESTIMATION FOR ELECTROMAGNETIC TRACKER USING A PRIORI INFORMATION
Publication Date: 2005-Aug-05
The IP.com Prior Art Database
In an embodiment, this invention applies the concept of using a priori information for improving the quality of initial seed. A specific seed calculation method for mobile C-arm navigation is disclosed here to prove the concept.
BACKGROUND OF THE INVENTION
 This invention discloses a two-step seed algorithm for improving robustness of the new tracking method. Since the two-step seed algorithm is essentially a numerical table-lookup based approach, it may not work effectively and reliably for the entire tracking space. The experimental results have shown that the iterative tracking algorithm may not always coverage to a stable position and orientation (P&O) solution for the tracker positions where the initial seeds significantly deviate from the actual sensor position. Usually the diverged P&O output can be detected by some kind of accuracy checking mechanism and removed from navigation data set to avoid patient injury. However, it does not restore the navigation capability at those locations. Thus, an alternative fix for inaccurate seed estimation is needed in addition to the proposed 2-step seed searching method.
DETAILED DESCRIPTION OF THE INVENTION
A priori information is very useful for solving inverse problems where the iterative parameter minimization methods are commonly used. For example, it is a popular approach in non-linear image reconstruction to use knowledge of anatomical properties to find high quality seeds for the algorithm .
In some specific navigation applications such as the mobile C-arm tracking system, the receiver is permanently attached to C-arm and moved together with the gantry on a well-controlled spatial path. To correlate the known C-arm motion information with the tracker data, we can largely improve the robustness of the tracking algorithm by having good initial guess of the receiver position to launch the fitter.
Using Bonneville 3D navigation system as an example, the design requires a highly repeatable motion control of the Flat Panel Detector (FPD) to sweep around the patient on the orbital plane. As illustrated in Figure 1, a receiver (Rx) pack is rigidly attached to the FPD, and a transmitter (Tx) is affixed to patient anatomy and remains the same location and pose with respect to the world coordinate system (WCS) during the 3D sweep. If we assume there is no C-arm deflection, we may fit the Rx positions to a perfect circle based on a subset of tracker data acquired during the C-arm sweep. The tracker data used for circle fitting should be those converged outputs from the field mapping program. The goal is to estimate the missing Rx positions that the fitter does not work because of inaccurate seeds.
To do this, a Gauss-Newton algorithm is used to find estimate of rotation and translation parameters that transform the data to the best-fit circle. As a result, we may obtain the rotation center (x0, y0, z0), radius R, and normal direction of the fitted circle. A patient coordinate system (CS) is then defined at the rotation center having the same axis directions as the Tx CS.
The next step are to express the Rx P&O in the pati...