Adaptive step size LMS for ECG artefact reduction during MRI

During cardiac MRI, fast switching gradients cause artifacts on the electrocardiogram (ECG), disturbing both triggering and patient monitoring. To cancel this noise, the Least Mean Squares (LMS) algorithm is a simple and efficient method. LMS uses one main parameter, its step size, which influences the quality of artifact reduction. We propose a method using the MR gradient variance to choose this parameter accurately using information about the sequence played by the MR scanner. The proposed method achieved systematically better results than the standard LMS with a 1.5T ECG database.