Adaptive removal of gradients-induced artefacts on ECG in MRI: a performance analysis of RLS filtering.

One of the main vital signs used in patient monitoring during Magnetic Resonance Imaging (MRI) is Electro-Cardio-Gram (ECG). Unfortunately, magnetic fields gradients induce artefacts which severely affect ECG quality. Adaptive Noise Cancelling (ANC) is one of the preferred techniques for artefact removal. ANC involves the adaptive estimation of the impulse response of the system constituted by the MRI equipment, the patient and the ECG recording device. Least Mean Square (LMS) adaptive filtering has been traditionally employed because of its simplicity: anyway, it requires the choice of a step-size parameter, whose proper value for the specific application must be estimated case by case: an improper choice could yield slow convergence and unsatisfactory behaviour. Recursive Least Square (RLS) algorithm has, potentially, faster convergence while not requiring any parameter. As far as the authors' knowledge, there is no systematic analysis of performances of RLS in this scenario. In this study we evaluated the performance of RLS for adaptive removal of artefacts induced by magnetic field gradients on ECG in MRI, in terms of efficacy of suppression. Tests have been made on real signals, acquired via an expressly developed system. A comparison with LMS was made on the basis of opportune performance indices. Results indicate that RLS is superior to LMS in several respects.