1338483 - Quantifying Cardiac Function in the Presence of Implantable Cardioverter Defibrillators with Cardiovascular Magnetic Resonance Imaging: Evaluation in Healthy Volunteers

Cardiovascular MRI (CMR) cine is the gold standard for ventricular mass and volume measurements and provides reference left ventricular ejection fraction (LVEF) values when compared to other imaging modalities [1]. Typical cine sequences use steady-state free precession (SSFP) sequences, which suffer from banding artifacts caused by metallic implants such as implantable cardioverter defibrillators (ICDs) or pacemakers [2]. As an alternative, high-bandwidth gradient-echo (GRE-HBW) cine sequences can be used to minimize device artifacts [3]. By increasing the bandwidth and reducing slice thickness, device artifact area is reduced with a slight SNR penalty (Fig. 1). Previous studies have assessed image quality of SSFP and GRE cines in the presence of ICDs [4] or compared left ventricular ejection fractions (LVEFs) between SSFP cines and GRE cines without ICDs [5]. In this study, we quantify LVEF in healthy volunteers with and without a taped device using GRE-HBW and SSFP cines, respectively. 

Methods:

11 healthy volunteers (average age: 26 years, seven males) were scanned on a research 1.5 Tesla scanner (MR450w, GE Healthcare, Waukesha WI), and all volunteers provided written informed consent. An ICD (Boston Scientific, Marlborough, MA) was taped to the volunteers’ left pectoral muscle approximately 2” below the clavicle and 2” medial to the shoulder, a representative location to clinical implantation locations. 

SSFP sequences used the following parameters: FOV=35 x 35cm2, resolution=1.4 x 1.4mm2, slice thickness/slice spacing = 5.2/4.0mm, bandwidth=651 Hz/px, flip angle=60°. GRE-HBW sequences used the following parameters: FOV=36 x 27cm2, resolution=2.1 x 2.2mm2, slice thickness/slice spacing = 4.0/4.0mm, bandwidth=1453.5 Hz/px, flip angle=20°. 

GRE-HBW and SSFP images were manually contoured by two readers (O1: CS, O2: XQ); all image analysis was conducted in CVI42 (Circle Cardiovascular Imaging, Calgary, Canada). For sample contours, see Fig. 2. LVEF, end systolic volume (ESV), end-diastolic volume (EDV), stroke volume (SV), and myocardium mass were calculated for each set of images.

Results:

No significant difference was observed between LVEF in SSFP and GRE-HBW cines (O1: -1.36%, p=0.60; O2: -2.29%, p=0.19). The Bland-Altman limits of agreement between GRE-HBW and SSFP LVEF are (-20.32%, +17.59%) and (-7.13%, +2.55%) for O1, O2 respectively (Fig. 3).  No significant differences were observed in ESV, EDV, SV, or myocardium mass as well. 

Feasibility of clinical auto-contouring software in GRE-HBW images was also investigated. From initial experience, auto-contouring is feasible when artifacts are far away from the endocardial border, but larger artifacts cause reduced epicardial volume and fewer slices contoured than human observers.

Conclusion:

In conclusion, we studied the differences in LVEF between SSFP and GRE-HBW cine with a device present. A good agreement is observed between SSFP and GRE-HBW LVEF values across multiple observers.