‘Image-navigated 3-dimensional late gadolinium enhancement cardiovascular magnetic resonance imaging: feasibility and initial clinical results’

Consecutive patients with known or suspected heart disease that underwent a CMR examination including LGE imaging at our facility were enrolled from February 2014 to February 2015. Patients were required to be ≥18 years of age and have no contraindications to gadolinium contrast, inclusive of an estimated glomerular filtration rate ≤ 60 ml/min/1.73 m². Patients with atrial fibrillation were excluded. Written informed consent was obtained from all participants. The institutional ethics board at Guy’s & St Thomas’ Hospital, King’s College London approved the study.

All participants were examined in supine position in a 1.5 Tesla CMR system (Gyroscan Intera, Philips Medical Systems, Best, The Netherlands) using a 32-element phased array cardiac synergy coil. The imaging protocol included cine imaging, standard 2D LGE imaging and iNAV-3D LGE imaging. 2D and iNAV-3D LGE image acquisitions were performed in random order to cancel out differences due to acquisition time after contrast agent injection.

Cine images were acquired using a ECG-gated balanced steady-state free precession based pulse sequence in serial short-axis slices from the atrioventricular annulus to the apex at 10 mm intervals, as well as in long-axis orientations (slice thickness 8 mm, gap 2 mm, echo time 1.5 ms, repetition time 3.0 ms, flip angle 50°). For LGE imaging, an intravenous bolus of 0.2 mmol/kg gadolinium (Gadovist®, Bayer Inc., Toronto, Ontario, Canada) was administered. High-resolution (2 mm isotropic) iNav-3D LGE and 2D LGE using T1-weighted RF-spoiled gradient-echo inversion recovery sequences were performed in random order 10 min following contrast administration. The image-navigator approach has been previously implemented for coronary CMR angiography [19], and was acquired using the ramp-up profiles of the 3D LGE sequence, as illustrated in Fig. 1. The ramp-up scheme consisted of linearly increasing amplitude of 10 RF pulses prior to the 3D LGE sequence. Phase encoding gradients were added to the ramp-up profiles to enable iNAV acquisition, using a high-low profile order. A high-low profile order was used to obtain higher signal for the central k-space line of the navigator, which was acquired with the highest flip angle, as well as closest possible temporal distance to the 3D LGE acquisition. As the ramp-up profiles were used to generate the iNAVs, size, location and orientation of the field-of-view for the iNAVs and 3D LGE were identical. To maximize sensitivity to detect respiratory motion for the iNAVs, the field-of-view was oriented in the coronal plane covering the whole heart, with readout in foot-head direction (2 mm resolution) and phase encoding in left-right direction (9 mm resolution). In the slice-encoding direction the iNAVs were projections of the field-of-view resulting in a slice thickness of approximately 100–120 mm. iNAV motion compensation was implemented in real-time on the scanner and has been previously described [20]. In brief, the location of the shim geometry over covering the heart was used to define a region-of-interest in the first iNAV, which was used as reference iNAV. The reference was registered to every subsequent iNAV using 2D normalized cross-correlation, yielding translational motion in foot-head and left-right direction. The estimated motion was used to perform translational correction of the 3D LGE raw data by applying a linear phase shift. Respiratory gating was implemented with constant respiratory efficiency using single end-expiratory threshold CRUISE algorithm [18]. This approach acquires twice as much data as needed to fill 3D LGE k-space (resulting in exactly 50% gating efficiency) and only the half acquired at the most end-expiratory was used to reconstruct the gated image. Data acquisition of the iNAV-3D LGE was ECG-triggered to the mid-diastolic rest period of every cardiac cycle. The 2D LGE was acquired with multiple slices, covering the entire left ventricle in short axis orientation, in addition to single slices in two-, three- and four-chamber views. The 2D LGE consisted of a phase sensitive inversion recovery (PSIR) protocol with inversion pulses every other RR interval. Separate Look-Locker scans were performed prior to iNAV-3D and 2D LGE, to find the optimal inversion times to null healthy myocardium for one and two RR-interval inversion pulse frequencies, respectively. The scan times were recorded for both iNAV-3D and 2D LGE. For the 2D LGE, the scan time was recorded as the time between the acquisition of the first and last slice, including pauses between breath-holds to capture the true scan time of all 2D slices.

Table 1 describes the scan parameters for each method.

Post processing was performed with dedicated software (Circle Cardiovascular Imaging 4.2, Circle Cardiovascular Imaging, Calgary, Canada) by two independent cardiologists specializing in CMR imaging (7-years experience in CMR/ SCMR Level 3 accredited and 3-years experience in CMR/ SCMR Level 2 accredited, respectively), who were blinded to the results of prior measurements by the same reader, measurements by other readers and all clinical data, as well as the order in which the 2D and iNAV-3D LGE studies had been acquired. All 2D and 3D data were anonymised and presented in random order.

3D LGE image data were reformatted to short-axis planes to correspond with the 2D LGE slices and optimise comparisons. On both the 2D and iNAV-3D LGE images, endocardial and epicardial contours were automatically placed by the software and reviewed by the reader. After segmentation, diagnostic performance and image quality data analysis of 2D LGE and iNAV-3D images was performed. The acquired images were analyzed with regard to the agreement of the global and segmental detection, pattern and transmurality of LGE for each segment of a 16-segment model of the myocardium proposed by the American Heart Association excluding the apex [21].

Quantitative image sharpness was calculated for the 2D and 3D iNAV-LGE datasets. For each patient, a short-axis slice with the least amount of scar and myocardial trabeculations was selected for sharpness analysis. The corresponding slice in the iNAV-3D LGE was reformatted to the same reconstructed resolution. In each image, 8 profiles were manually selected along the endocardial border, at equidistant locations, as shown in Fig. 2. The sharpness was defined as the distance in pixels between 20% and 80% of the pixel intensity range of the profile, and a lower pixel distance indicate a sharper border. The sharpness from the 8 profiles was averaged for each dataset and patient. To ensure measurements were comparable between patients, all images were reformatted to the same in-plane resolution of 0.67 mm.

SNR was calculated for 2D LGE and iNAV-3D LGE for blood (SNR_blood) and myocardium (SNR_myo). The analysis was performed by drawing region of interests in the left ventricular blood pool and septum of the myocardium to obtain signal from the respective tissues. Noise was defined as the standard deviation of a region of air in the lungs (SD_air), avoiding pulmonary vessels. The SNR values were calculated as the respective signal divided by SD_air, multiplied by a factor of 0.655 to account for the Rayleigh distribution of the noise [22]. Furthermore, blood-myocardium contrast-to-noise (CNR_blood-myo) was calculated using the same regions-of-interest as for the SNR calculations, subtracting SNR_myo from SNR_blood, dividing by SD_air and multiplying by 0.655 to account for the noise distribution.

To obtain a head-to-head direct comparison between the two sequences only complete and diagnostic sequence sets were used for further analysis.

At first, visual perusal of the images was performed to determine whether infarction was present or absent. Positive finding in terms of a detectable scar in the LGE images was defined as a visible late enhancement. Global and segmental LGE detection were subsequently characterized in terms of LGE pattern (subendocardial/ ischemic, subepicardial, patchy, diffuse, involving right ventricular insertion points). In datasets with discrepancy between the two techniques, the results were compared against a panel standard. LGE transmurality was visually assessed by the reader as the segmental spatial extend of LGE within each segment and graded by using a five-point scale (0: 0%; 1: 1%–24%; 2: 25%–49%; 3, 50%–74%; 4, 75%–100%; 5, striae; 6- diffuse).

LGE was detected in 5 patients with ischemic pattern (23%), in 7 with non-ischemic pattern (32%), while it was absent in 10 cases (45%). There was excellent agreement between 2D and 3D data sets with regard to global LGE detection. There were no significant differences between the two techniques in segmental LGE detection and transmural extent (P = N.S.). The 95% confidence intervals and global tests showed no significant differences in the LGE pattern among all the considered combinations (P = N.S.). Similar observations were noted when the effect of the order of each sequence acquisition was considered.

Table 2 summarizes the results of diagnostic quality assessment. Typical case examples in patients with LGE are provided in Fig. 3.

The mean blood-myocardium sharpness ± standard deviation was 8.5 ± 3.6 for 2D LGE and 9.4 ± 3.0 for iNAV-3D LGE. No statistically significant differences were found between 2D LGE and iNAV 3D LGE for the blood-myocardium sharpness measurements (P = N.S.). Examples from two patients with improved image sharpness obtained using iNAV-3D LGE compared to 2D LGE are shown in Fig. 4.

The SNR_blood, SNR_myo and CNR_blood-myo for 2D LGE and iNAV-3D LGE are summarised in Table 3. Statistically significant differences were found for SNR_blood (P < 0.001) and CNR_blood-myo (P < 0.001), showing higher SNR_blood and CNR_blood-myo for 2D LGE, while SNR_myo was not statistically significant between 2D LGE and iNAV-3D LGE.

All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.