A Kalman filtering framework for prospective motion correction

Introduction: Prospective motion correction is becoming a feasible means of overcoming the problem of patient motion in brain imaging. Navigator data from a tracking system is used to update the scanner gradients, and therefore the position of the imaging volume, before every spin excitation. Regardless of the tracking system used, accuracy is a critical factor as any position noise results in image artefacts [1]. Thus, position errors must be minimised. Due to latency in the system, one should also predict the position of the object at a certain time, , t Δ in advance. Finally, an estimate of residual position errors would be useful for post-processing correction. The goal of this work was to develop a pose prediction method based on Kalman filtering that reduces tracking noise and enables retrospective estimation of residual errors. The method was validated using simulated and experimental pose data from an optical tracking system (similar to that described in [2]).