Motion artifact reduction in magnetic resonance imaging through navigator processing

Motion remains one of the biggest obstacles to reliably producing high-quality magnetic resonance (MR) images. MR navigators are used to provide one-dimensional image profiles of the anatomy of interest. By comparing pairs of navigator profiles, translational shifts can be calculated and used in techniques to compensate for motion artifacts. Existing motion correction techniques can be negatively impacted by external factors, such as changes in motion patterns and high noise. Improved navigator processing can address these issues.;Shift calculations can be used in real-time applications to prospectively determine which data to reacquire with the goal of converging onto a target shift value for all data. By filtering the shift histograms, the initial target calculation can be improved. Recalculating the target throughout the reacquisition period improves resistance to changes in motion patterns. Depending on the anatomy being imaged, inconsistently shaped navigators can occur and cause misleading shift calculations. By incorporating navigator integrity into the decision-making process, data acquisitions corresponding to these navigators can be replaced. If a uniquely shaped navigator is chosen as the reference, all subsequent shift calculations can be misleading. By calculating a profile from a set of navigators, a reliable reference is formed and shift calculation quality is generally preserved.;The accuracy of a shift calculation depends significantly on the signal-to-noise ratio of the navigator data. By applying a self-filter to a navigator, noise can be balanced. Since the energy of a navigator is often concentrated at the lower frequencies, self-filters can have the negative effect of broadening navigator profiles and subsequent correlation peaks, thus lowering shift calculation accuracy. Alternative shift calculation methods using phase-only matched filters, complement-N filters and locally nonlinear matched filters can be effectively used to maintain shift calculation accuracy in the presence of noise after navigator profiles have been self-filtered. A more effective technique uses the average navigator magnitude and estimates the noise to dynamically create a Wiener filter as the optimal linear filter for navigator noise removal.