Introduction: Patients with a cardiac implantable electronic device (CIED) have a high burden of arrhythmia, in whom standard ECG-gated breath-held (BH) cine MRI with gradient-recalled echo (GRE) readout often produces poor or non-diagnostic image quality. In such patients, a free-breathing, real-time (FB-RT) cine MRI with GRE readout may be desirable. We had previously described a 32-fold accelerated FB-RT cine using a combination of radial k-space sampling, view-sharing (VS), and k-space weighted image contrast (KWIC) filtering for imaging patients with CIEDs [1]. The purpose of this study was to determine whether the proposed pulse sequence produced diagnostically acceptable image quality and clinically acceptable bi-ventricular volumes and functional parameters in CIED patients.

Methods: MRI: We enrolled 20 patients (8 females; mean age = 64.44 ± 16.44 yrs) with a CIED (see Table 1A for demographics) and performed two different cine scans on a 1.5T MRI scanner (Avanto,Siemens Healthineers): clinical BH cine and the proposed FB-RT [1]. The relevant imaging parameters are summarized in Table 1B.

Image Reconstruction: For FB-RT cine, images were constructed on a GPU (2 x 32GB Tesla V100) workstation using the iterative Golden-angle Radial Sparse Parallel (GRASP) [4] framework adapted to reconstruct 6 radial spokes per frame supplemented by 18 shared spokes with KWIC [2] filtering to remove the center of shared k-space lines, as shown in Figure 1.

Image Analysis: LV and RV cardiac contours were segmented using the automatic AI tools followed by manual correction in Circle (cvi42,v5.13.9) for BH and FB-RT cine images. For FB-RT with multiple heartbeats per slice, we analyzed the representative (not short or long) heartbeat. Bi-ventricular volumes and functions were quantified. Bland-Altman analysis was conducted to determine the level of agreement between BH cine with FB-RT cine. Two attending cardiologists visually evaluated the image quality (randomly subsampled to 14 patients) using a 5-point Likert scale for each of four categories as shown in Table 2, where a score of 3 defines clinically acceptable. Summed visual score (SVS) is defined as the numerical sum of all scores, and SVS > 12 is defined as clinically acceptable.

Results: Figure 2 shows representative three short-axis images of one patient in sinus rhythm (CV in R-R interval = 1.88%; see caption in Table 1 for definition of CV) and another patient in atrial fibrillation (CV in R-R interval = 33.22%). While BH cine produced ghosting artifacts in patients with arrhythmia, FB-RT cine produced good image quality in all patients. As shown in Figure 3, both the left ventricle (LV) and right ventricle (RV) parameters were in good agreement; mean difference ranged from -1.6 (-2.1% of mean) to 0.6 (0.9% of mean) for the LV and 0.3 (0.7% of mean) to 1.5 (1.2% of mean) for the RV; there was no systematic trend in any of the Bland-Altman plots. None of the functional parameters were significantly different (p>0.1). As summarized in Table 2, none of the individual image quality scores were significantly different (p>0.32) except for noise; however, all four individual scores were > 3.0 and SVS > 12 for FB-RT radial cine.

Conclusions: Our 32-fold accelerated radial FB-RT cine produced diagnostically acceptable image quality and relatively accurate bi-ventricular volumes and functional parameters compared with standard BH cine in patients with a CIED. This pulse sequence may be useful for shortening overall scan time and producing diagnostically useful images in all CIED patients, even those with arrhythmias and/or dyspnea.

Acknowledgements

The authors would like to thank funding support from the National Institutes of Health (R01HL116895, R01HL151079, R21EB030806A1, 1R01HL167148‐01A1), the American Heart Association (19IPLOI34760317, 949899), and the Radiological Society of North America (EILTC2302).