Filtering of ECG signals distorted by magnetic field gradients during MRI using non-linear filters and higher-order statistics.

The electrocardiogram (EGG) is the state-of-the-art signal for patient monitoring and gating in cardiovascular magnetic resonance (CMR) imaging applications. However, EGG signals are severely distorted during MRI scans due to the effects of static magnetic fields, radio frequency pulses and fast-switching gradient magnetic fields. Gradient-induced artifacts that cause high frequency peaks in the EGG signal especially hamper a correct and reliable QRS detection. To cope with this problem, a new median-based real-time gradient filter (M1) approach was developed. To improve the filter results, a preprocessing step based on higher-order statistics (M2) was added to this. For the evaluation of the filtering techniques, EGG signals were acquired in a 3T MRI scanner during different MR sequences. A qualitative comparison was made using the mean square error as well as the signal power before and after filtering and the results of the QRS detection. Here, reliable results were achieved (detection error rate [DER] M1: 0.23%, HER M2: 0.74%). It was shown that the two developed techniques allowed a reliable suppression of the gradient artifacts in real time.