Target Audience Neurologists, radiologists and scientists interested in obtaining high-fidelity 12-lead ECGs for physiological monitoring during diagnostic or interventional brain functional (fMRI) or Diffusion-weighted (DW-EPI) EPI studies.

Purpose ECG traces inside MRI bore are distorted by both the magneto-hydrodynamic effect and by strong (~5V) MRI gradient- induced electric fields. We previously developed and validated [1] a prototype MRI-compatible 12-lead ECG platform that utilized a theoretical equation with 19 parameters to remove gradient-induced-voltage (GIV) artifacts created during high-gradient-duty-sequences (SSFP, FSE, GRE) on 12-lead ECGs. The system acquired the 12-lead ECG traces, along with the three gradient waveforms that produce the GIV. A training procedure (duration 3-6 seconds) acquired ECGs during periods without gradients, followed by parallel-imaging accelerated sequences, to determine the 19 equation parameters. Thereafter, GIVs were removed in real-time from ECGs acquired during the imaging sequences. When the 12-lead ECG electrodes are not positioned close to magnet iso-center, such as during brain imaging, the electric fields induced in the torso are larger [2], so the GIVs induced on ECGs traces can be 3-4 times larger than observed in cardiac/abdominal imaging. Additionally, during DW-EPI’s diffusion-encoding the highest gradient strengths are employed, while during both fMRI and DWI read-out gradients are modulated rapidly at the scanner’s maximal slew rate, producing larger GIVs than other sequences. In this study, we test the ability of the theoretical method [1], together with an improved acquisition system, to remove GIVs in this most-difficult case. DW-EPI and fMRI are used clinically to diagnose patients undergoing acute stroke and brain seizures in MRI-equipped emergency rooms, as well as detect tumor and fiber location during MRI-guided brain surgery, so the availability of on-line diagnostic-quality 12-lead ECG monitoring can be critical. Additionally, ECG-gating is increasingly used to acquire more repeatable EPI-based diffusion and perfusion parameters.

Methods A theoretical equation for the gradient-induced voltages on each ECG electrode was derived [1], which depends on the three gradients and their time derivatives. It has 19 terms, and requires the determination of 19 parameters. The ECG traces were collected using a commercial GE (Waukesha, WI) CardioLab system. In this study, we used the Cardiolab catheter input channels, instead of the surface ECG channels [1], thus preventing Cardiolab’s Right-Leg voltage drive from sending out strong voltages to the subjects, and enabling measurement of the GIVs induced on each limb lead independently. We used an amplification value of 50 to prevent amplifier saturation, high-pass filters of 0.5 Hz to prevent respiratory motion and 1st-order low-pass filters of 150 Hz. The training protocol was identical to that used previously [1], acquiring ECG traces during GRAPPA=5-6 accelerated single-slice EPI sequences, followed by acquisition of ECGs during 3 QRS cycles without gradient activity. Training determined the 19 equation parameters. Since some strong GIV spikes remained in the training data after “cleaning”, we developed a method to remove them by Fourier-transforming the cleaned traces, determining the spike characteristic frequencies (all >50 Hz), and designing notch filters to remove signals at those frequencies whose amplitude was above a threshold (~15mV). Full resolution multi-slice EPI imaging (GRAPPA=2) was performed after the training protocol, with real-time subtraction of the computed gradient-induced voltages based on the 19 coefficients and the simultaneously recorded gradient waveforms, followed by the notch IIR filters. Healthy volunteer images were acquired on a 3T Siemens Skyra (Gradients: 45mT*m-1, 200 T*m-1*sec-1 slew rate). fMRI sequence parameters: TR/TE/flip=700ms/51ms/900, 102x128,epi factor=51,GRAPPA=2, 27 cm fov, 6 mm slice, 10 slices/TR, BW=2170 Hz/pxl, DW-EPI parameters: B==0 and1000 s1mm-2, TR/TE/flip=800ms/101ms/300, 192x192,epi factor=71, GRAPPA=2, 20 cm FOV, 6 mm slice, 6 slices/TR, BW=2170Hz/pxl. 12-lead ECG electrodes were placed at standard torso locations.

Results Figure 1 shows accelerated (training) and the dual-echo DW-EPI images. Figure 2A shows 9 independent ECG traces (Precordial leads:V1-V6, Limb leads: Left-Arm, Right-Arm, Left-Leg), acquired during an fMRI sequence, along with the simultaneously-acquired X, Y and Z gradient waveforms, before and after GIV removal. Figure 2B is a magnified view of 4 ECG traces. Raw traces with GIV (red), after cleaning with the equation (Green), and after cleaning with the equation followed by the IIR filter (Blue). >95% of the GIV voltage is removed. Some residual GIV remains at the start and end of each imaging period, due to the amplifier’s non-linear response to the strong GIV signal.