Accuracy of self-navigated free-breathing isotropic 3D whole heart inversion recovery magnetic resonance for detection of myocardial infarction

Background

Cardiac magnetic resonance (CMR) allows detection of myocardial scar after myocardial infarction. Usually 2D image planes in short-axis and three long axis orientations are obtained. However to plan in patients with scar e.g. complex electrophysiological intervention for reentry arrhythmias, high-resolution 3D information of the scar is highly desirable. This study therefore evaluates the accuracy of self-navigated isotropic 3D-free-breathing CMR with inversion recovery (3D-SNIR) to detect myocardial scar tissue.

Methods

Patients after myocardial infarction detected by late gadolinium enhancement on standard 2D inversion recovery sequences (2D LGE) underwent a CMR exam with 3D-SNIR on a 1.5T clinical CMR scanner (Aera, Siemens, Germany). Data acquisition was performed during the most quiescent systolic phase with a prototype segmented 3D radial trajectory with self-navigation after administration of 0.2mmol/kg of Gadobutrol. A non-selective IR pulse was added prior to each acquired k-space segment to the segmented, ECG-triggered, fat-saturated radial SSFP imaging sequence. Parameters: TR/TE 3.1/1.56ms, FOV (220mm)³, matrix 192³, isotropic voxel size (1.15mm)³, RF excitation angle 115°, and receiver bandwidth 898Hz/Px. TI (= 250-300ms) was assessed with a 2D radial scout scan prior to 3D-SNIR. A total of about 12'000 radial readouts were acquired for each 3D scan during free breathing with 100% respiratory efficiency. 3D LGE datasets were compared to standard 2D LGE for scar tissue detection with Osirix® software. Short axis 10mm slices were reconstructed from 3D LGE datasets by maximum intensity projection to yield a slice thickness of 10mm. Scar tissue was segmented on reconstructed slices on standard 2D and 3D-SNIR LGE and multiplied by slice thickness.

Results

Thirteen patients (5 females, age 58±10y) were included. Time between 2D LGE and 3D LGE was 59 ± 64 days. 3D-SNIR successfully corrected for respiratory motion in all acquisitions. All scars visualized by 2D LGE could be identified by 3D-SNIR (example see figure 1). Bland Altman-analyses and correlations showed a good agreement of quantification of scar volume obtained by 3D-SNIR compared to standard 2D LGE: -6.3±4.1ml, linear regression: r=0.977, p<0.001 (figure 2). 3D scar volume was 24.3±15ml vs. 2D 30.6±17ml, p<0.001. Intraobserver variability was 0.3±4.9ml, r=0.985, p<0.001; interobserver 1.5±9.9ml, r=0.74, p=0.014.

Figure 1

Examples for detection of myocardial scar in the septal (figure 1A) and inferior wall (figure 1B) by 2D (left) and 3D (right) late gadolinium enhancement in 2 patients.

Figure 2

Correlation of volume of scar determined by 2D and 3D late gadolinium enhancement

Conclusions

Detection of myocardial scar by 3D-SNIR is feasible and shows a good agreement with standard 2D LGE. The mean difference of -6.3ml might be explained by the higher spatial resolution of the 3D sequence. Integration of a phase sensitive inversion recovery pulse warrants testing to further improve 3D quantification of scar.

Funding

N/A.