Intra-MRI extraction of diagnostic electrocardiograms using dynamic feedback from carotidal magnetohydrodynamic voltages

Background

During Cardiac Magnetic Resonance Imaging (CMR), blood plasma electrolytes ejected into the aorta during early systole interact with the static magnetic field of the MR scanner (B₀) to produce a Magnetohydrodynamic (MHD) Effect [1]. Electrocardiograms (ECGs) recorded in the presence of B₀ are overlaid with induced MHD voltages (V_MHD), leading to non-robustly synchronized imaging [2], and preventing reliable physiological monitoring inside the MRI [3]. Previous methods have sought to separate between V_MHD and the true ECG (ECG_real) through adaptive filtering [3], independent component analysis [4], and advanced computational models [5]. However, these methods are based on a static model, which has limited accuracy during varying-rate heart-beats. We aim to develop accurate ECG_real extraction, as well as real-time Stroke Volume (SV) estimation (proportional to the integral of MHD over systole) [6], with the advantage of physiological feedback through the real-time monitoring of common carotidal MHD, through which the previously static MHD template can be dynamically updated, providing an increased level of accuracy during variations in heart rate, and a continuous estimation of V_MHD and ECG_real, for the patient's entire duration inside the MRI.

Methods

12-lead ECGs were acquired in two (n = 2) healthy volunteers during 20-second breath-holds in a 3T MRI (Figure 1ab) with the heart at isocenter. A secondary monitor was used to acquire a single anterior-posterior bipolar lead placed approximately on the left common carotid artery (Figure 1c). ECGs were acquired inside (ECG_real + V_MHD) and outside (ECG_real) the MRI bore during an initial phase in which a static MHD template was extracted, based on lead subtraction. Carotidal MHD was extracted from the single bipolar lead and phase-compensated to match V_MHD obtained from the 12-lead ECG. Carotid MHD was subsequently used to adaptively train a Least Mean Squares filter (Figure 1d) to update the MHD template and produce: (1) clean 12-lead ECGs; and (2) an accurate SV estimate [6] (Figure 1e).

Figure 1

Active removal of induced Magnetohydrodynamic voltages in ECGs recorded inside a 3T MRI using adaptive filtering.

Results

The adaptive filtering method was shown to reduce V_MHD in the acquired 12-lead ECGs, with residual noise forming <5% of the R-wave amplitude. The method preserved the true S-T segment, while requiring only a short training phase for the 12-lead ECG (10-15 seconds). The Pearson's Correlation Coefficient between Aortic and Carotid MHD increased from 0.51 to 0.88 after the adaptive filtering routine was applied. Figure 1f shows the extracted 12-lead ECG acquired inside the MRI bore after the training phase.

Conclusions

A method to extract true sinus rhythm beats from intra-MRI 12-lead ECGs was presented and shown to provide accurate dynamic measurements of induced V_MHD using Carotid artery MHD and ECG_real to allow for advanced physiological monitoring inside the MRI.