Abstract
Purpose
To develop an accelerated three-dimensional (3D) late gadolinium enhancement (LGE) pulse sequence using balanced steady-state free precession readout with stack-of-stars k-space sampling and extra motion-state golden-angle radial sparse parallel (XD-GRASP) reconstruction and test the performance for detecting atrial scar and fibrosis in patients with atrial fibrillation (AF).
Materials and Methods
Twenty-five patients with AF (20 paroxysmal and five persistent; 65 years ± 7 [standard deviation]; 18 men) were imaged at 1.5 T using the proposed LGE sequence with 1.3 mm × 1.3 mm × 2-mm spatial resolution and predictable imaging time. The resulting images were compared with historic images of 25 patients with AF (18 paroxysmal and seven persistent; 67 years ± 10; 14 men) obtained using a reference 3D left atrial (LA) LGE sequence with 1.3 mm × 1.3 mm × 2.5-mm spatial resolution. Two readers visually graded the 3D LGE images (conspicuity, artifact, noise) on a five-point Likert scale (1 = worst, 3 = acceptable, 5 = best), in which the summed visual score (SVS) of 9 or greater was defined as clinically acceptable. Appropriate statistical analyses (Cohen κ coefficient, Mann-Whitney U test, t tests, and intraclass correlation) were performed, where a P value < .05 was considered significant.
Results
Mean imaging time was significantly shorter (P < .01) for the proposed pulse sequence (5.9 minutes ± 1.3) than for the reference pulse sequence (10.6 minutes ± 2). Median SVS was significantly higher (P < .01) for the proposed (SVS = 11) than reference (SVS = 9.5) 3D LA LGE images. Interrater reproducibility in visual scores was higher for the proposed (κ = 0.78–1) than reference 3D LA LGE (κ = 0.44–0.75). Intrareader repeatability in fibrosis quantification was higher for the reference cohort (intraclass correlation coefficient [ICC] = 0.94) than the prospective cohort (ICC = 0.79).
Conclusion
The proposed 3D LA LGE method produced clinically acceptable image quality with 1.5 mm × 1.5 mm × 2-mm nominal spatial resolution and 6-minute predictable imaging time for quantification of LA scar and fibrosis in patients with AF.
Supplemental material is available for this article.
© RSNA, 2020
Summary
A three-dimensional left atrial late gadolinium enhancement pulse sequence using balanced steady-state free precession readout with stack-of-stars k-space sampling and extra motion-state golden-angle radial sparse parallel reconstruction was developed and tested to determine whether it was capable of producing clinically acceptable image quality with high nominal spatial resolution (1.5 mm × 1.5 mm × 2 mm) and predictable imaging time (6 minutes) for quantification of left atrial fibrosis in patients with atrial fibrillation at 1.5 T.
Key Points
■ The mean imaging time was significantly shorter (P < .01) for the proposed pulse sequence (5.9 minutes ± 1.3) than for the reference pulse sequence (10.6 minutes ± 2).
■ The average-reader median summed visual score (SVS) (3 = nondiagnostic; 9 = clinically acceptable; 15 = excellent) was significantly higher (P < .01) for the prospective cohort images obtained using the proposed pulse sequence (SVS = 11) than the retrospective cohort images obtained using the reference pulse sequence (SVS = 9.5).
■ Interrater reproducibility in visual scores was higher for the prospective cohort images (κ = 0.78–1) than the retrospective cohort images obtained using the reference pulse sequence (κ = 0.44–0.75).
Introduction
Atrial fibrillation (AF) is the most common sustained arrhythmia in adults in the United States (1). Catheter ablation is superior to antiarrhythmic drugs for rhythm control treatment of AF, but success rates to date are still modest: approximately 70% for paroxysmal AF and 60% or lower for persistent AF (2). Identifying accurate predictors for AF recurrence based on clinical (eg, AF duration) and imaging metrics (eg, left atrial [LA] size or left ventricular function) has proven difficult. A more promising candidate for predicting AF recurrence following catheter ablation is LA fibrosis because atrial fibrosis has been shown to play a crucial role in the development of the arrhythmogenic substrate for AF and may be a marker for more extensive disease less amenable to pulmonary vein isolation (3,4). Although LA fibrosis as assessed with three-dimensional (3D) LA late gadolinium enhancement (LGE) MRI (5) has shown promise for predicting AF recurrence following catheter ablation (6–8), the “Utah” classification (9) has not been independently reproduced by the field, most likely the result of several major gaps in technology: (a) inadequate spatial resolution (1.5 mm × 1.5 mm × 2.5 to 5 mm) (5,6,8,10,11), (b) lengthy imaging time (approximately 10–15 minutes) (12), and (c) unreliable image analysis techniques for quantification for LA fibrosis in the thin (1–2 mm) LA wall. These gaps prevent widespread adoption of LA fibrosis quantification.
We sought to address the four limitations (spatial resolution, imaging time, reliance on external navigator gating, and reliance on 3 T) of standard 3D LA LGE MRI by synergistically combining the following advanced techniques: (a) balanced steady-state free precession (SSFP) readout (higher signal-to-noise ratio [SNR] than gradient echo), (b) stack-of-stars k-space sampling (higher incoherence than Cartesian k-space sampling), and (c) self-navigation of respiratory motion using the extra motion-state golden-angle radial sparse parallel (XD-GRASP) MRI (13) framework to achieve 100% gating efficiency and predictable imaging times. The purpose of this study was to develop a 3D LA LGE MRI pulse sequence integrating the aforementioned techniques and test whether it was capable of achieving clinically acceptable image quality with high spatial resolution (1.5 mm × 1.5 mm × 2 mm) and predictable imaging time (approximately 6 minutes) in study participants with AF at 1.5 T. In addition, the results of this study were compared with historic data obtained using the reference 3D LA LGE pulse sequence with 1.3 mm × 1.3 mm × 2.5-mm spatial resolution and approximately 11-minute imaging time (6).
Materials and Methods
Patients
We prospectively enrolled 25 study participants (18 men, seven women, mean age = 65 years ± 7 [standard deviation], 20 paroxysmal and five persistent AF) undergoing a clinically indicated preablation MRI with 0.20 mmol/kg of gadobutrol (Gadavist; Bayer Healthcare Pharmaceuticals, Whippany, NJ) from January 2019 to February 2020. Six of the 25 study participants underwent prior pulmonary vein isolation alone; two of whom underwent radiofrequency ablation; four of whom underwent cryoablation. Our 3D LA LGE MRI examinations were conducted at 19 minutes ± 3 after administration of contrast agent, in which this timing was calculated as the difference in time stamp between the contrast material–enhanced MR angiography and 3D LA LGE acquisitions. This study was conducted in accordance with protocols approved by our institutional review board and was compliant with the Health Insurance Portability and Accountability Act; all participants provided informed consent in writing. See Table for relevant clinical profiles.
Patient Demographics and Clinical Profiles
For comparison with historic data, we retrospectively identified 25 patients (14 men, 11 women, mean age = 67 years ± 10, 18 paroxysmal and seven persistent AF) at a second institution who had undergone reference 3D LA LGE MRI (6) with 0.10 mmol/kg of gadobenate dimeglumine (MultiHance; Bracco, Milan, Italy) from March 2016 to December 2016. Retrospective patients were selected to match demographics of the prospective cohort. Nearly matching patient characteristics included age, AF type, AF duration, prior cardioversion and AF ablation, coronary artery disease, and CHA2DS2-VASc score. Seven of the 25 patients underwent prior pulmonary vein isolation plus posterior wall debulking using radiofrequency ablation. Reference 3D LA LGE MRI examinations were conducted at 23 minutes ± 5 after administration of contrast agent. The need for informed consent was waived. A retrospective study was approved by the institutional review board of the second institution and was found to comply with the Health Insurance Portability and Accountability Act. See Table for relevant clinical profiles.
MRI Hardware
The MRI examinations of the prospective study participants were conducted with two 1.5-T whole-body MRI scanners (MAGNETOM Aera and Avanto; Siemens Healthcare, Erlangen, Germany) equipped with a gradient system capable of achieving a maximum gradient strength of 45 mT/m and a maximum slew rate of 200 T/m/sec. The radiofrequency excitation was performed with the body coil, and the signal reception was made with the body matrix and spine coil arrays (approximately 18 elements for Avanto and approximately 30 elements for Aera). The MRI examinations in the retrospective patients were also conducted with a 1.5-T scanner (Avanto) equipped with identical specifications, but at the second institution.
Pulse Sequence
Rationale for balanced SSFP readout and stack-of-stars k-space sampling with unique inversion are described in the Appendix E1 (supplement). The proposed 3D LA LGE MRI was performed using the following parameters: image acquisition matrix size = 192 × 192 × 48, field of view = 288 mm × 288 mm × 96 mm, nominal spatial resolution = 1.5 mm × 1.5 mm × 2 mm, receiver bandwidth = 704 Hz/pixel, echo time/repetition time = 1.8/3.5 msec, flip angle = 40°, inversion time to null the normal myocardium = 220–350 msec, a stack of 50 k-space rays (48 for data + two for navigator) acquired per shot (heartbeat), electrocardiographic (ECG) triggering every heartbeat, readout duration per heartbeat = 168 msec, imaging time = 350 heartbeats, and oblique sagittal image orientation (Fig 1). Optimal inversion time was determined visually based on images obtained using an inversion time scout sequence with ECG triggering every heartbeat.
Figure 1:
A, Pulse sequence timing diagram of three-dimensional (3D) late gadolinium enhancement (LGE) pulse sequence using b-SSFP readout. Following an inversion pulse, ramp-up pulses are played to quickly approach the steady state of magnetization. Immediately before the readout, two navigator echoes are sampled, with at least one oriented along the head-to-foot direction. B, The 3D volume is oriented in oblique sagittal direction to sample the left side of the heart. C, “Winding staircase”–like k-space ordering with TGA sequence along kz and GA sequence along heartbeat. In this acquisition scheme, a unique inversion time (TI) samples the center of k-space, which is ideal for inversion-recovery–based LGE. b-SSFP = balanced steady-state free precession, GA = golden angle, TGA = tiny golden angle.
For the reference 3D LA LGE MRI sequence using gradient-echo readout (6), the relevant scan parameters included: image acquisition matrix size = 288 × 288 × 44–48, axial field of view = 360 mm × 360 mm × 110–120 mm, nominal spatial resolution = 1.3 mm × 1.3 mm × 2.5 mm, receiver bandwidth = 289 Hz/pixel, echo time/repetition time = 2.2/5.1 msec, flip angle = 19°, 0.875 partial Fourier factor in the phase-encoding direction, 0.8 partial Fourier factor in the partition-encoding direction, generalized autocalibrating partially parallel acquisitions (GRAPPA) (14) with reduction factor of two in the phase-encoding (right-left) direction and 52 autocalibration lines, imaging time = 6–15 minutes, 27–31 k-space lines per shot with readout duration per heartbeat = 138–158 msec, and axial image orientation. To minimize partition-encoding wrap artifact along the partition direction, we applied the following slice oversampling: 10% for 40 slices, 9.1% for 44 slices, and 8.3% for 48 slices. Inversion recovery preparation was applied every heartbeat, and fat saturation was performed immediately before data readout. The navigator was positioned on the right hemidiaphragm and navigator acceptance window was ±3 mm. Optimal inversion time was determined visually based on images obtained using an inversion time scout sequence with ECG triggering every heartbeat.
Image Reconstruction
Figure 2 shows the XD-GRASP image reconstruction pipeline used in this study. Image reconstruction was performed offline on a GPU workstation (Tesla V100 with 64GB memory, NVIDIA, Santa Clara, Calif; 32 Xeon E5–2620 v4 128 GB memory, Intel, Santa Clara, Calif) equipped with MATLAB (R2017b, MathWorks, Natick, Mass) running on a Linux operating system (Ubuntu16.04). For details on image reconstruction, see Appendix E1 (supplement).
Figure 2:
XD-GRASP reconstruction pipeline. The raw k-space data underwent gradient delay correction using the radial intersections (RING) method. In addition, from the raw data, the navigator ray oriented along the head-to-foot direction was used to extract respiratory motion. Using the gradient delay corrected data in combination with the navigator, each of six respiratory bins was populated with the same number of k-space data by adapting the bin width. In a separate preprocessing step in Cartesian coordinates, the coil sensitivities (variable S) were calibrated intrinsically from the non–motion-resolved images as shown. Next, the zero-filled, multicoil, motion-resolved images (variable x) were reconstructed using GPU NUFFT along with coil sensitivities as shown. In the compressed sensing reconstruction step, we used TTV and TPCA as the two orthogonal sparsifying transforms along the respiratory dimension, and nonlinear conjugate gradient with back-tracking line search as the optimization algorithm with 30 iterations. F represents undersampled FFT, S represents coil sensitivities, x represents image to be reconstructed, y represents the k-space data, T represents the finite difference operator, α represents the normalized regularization weight, and β represents the normalized fidelity weight. FFT = fast-Fourier transform, GPU = graphics processing unit, NUFFT = nonuniform fast-Fourier transform, TPCA = temporal principal component analysis, TTV = temporal total variation, XD-GRASP = extra motion-state golden-angle radial sparse parallel.
Visual Analysis of Image Quality
To minimize the impact of temporal blurring induced by compressed sensing, only the end-expiration frame was analyzed for this study (see Movie 1 [supplement] for dynamic display of respiratory motion-resolved, zero-filled and compressed sensing reconstructed images). Two cardiologists (B.D.W. and M.I.) with 15 and 3 years of clinical experience with cardiac MRI, respectively, graded the image quality. In total, 50 3D LGE image sets (25 sets each for the proposed and reference 3D LA LGE) were randomized and de-identified for display on a DICOM viewer (RadiAnt DICOM Viewer; Medixant, Poznan, Poland). Prior to visual evaluation, the readers were given training data sets with poor to excellent quality to calibrate readers’ scores in consensus. Following this training session, the readers independently graded the images by being blinded to each other’s score and the clinical history of patients. The readers were not completely blinded to image acquisition type because this information was easily discernable as a result of differences in image orientation (oblique sagittal for the proposed vs axial for the reference) (see Fig 3). Consideration was given to reformat the proposed data to axial or the historic data to oblique sagittal but was not performed because reformatting leads to degradation in image quality because of the nonlinear process. As such, we decided to evaluate the images in their native orientation for fairness. The readers graded for each of three categories on a five-point Likert scale: conspicuity of LA wall (1 = nondiagnostic; 2 = poor; 3 = clinically acceptable; 4 = good; 5 = excellent), noise throughout (1 = nondiagnostic; 2 = severe; 3 = moderate; 4 = mild; 5 = minimal), and any visible artifact on the LA (1 = nondiagnostic; 2 = severe; 3 = moderate; 4 = mild; 5 = minimal). In addition, a summed visual score (SVS) was computed as the sum of each category (conspicuity of LA wall, noise, and artifact) to represent the overall image quality. A score of 3 was defined as clinically acceptable for each individual category, whereas a score of 9 was defined as clinically acceptable for SVS.
Figure 3:
Representative best, average, and worst images for the proposed (top row) and reference (bottom row) three-dimensional (3D) left atrial (LA) late gadolinium enhancement (LGE) acquisitions. Arrows point to LA wall. For dynamic display through slice, see Movies 2–7 (supplement).
Movie 1:
Video display of respiratory motion-resolved zero-filled (left) and compressed sensing reconstructed (right) images.
Quantification of LA Fibrosis and Volume
One reader (S.G.) with 3 years of experience as a graduate student in cardiac MRI quantified LA fibrosis on both the reference and prospectively acquired data using commercial software (ADAS 3D; Galgo Medical, Barcelona, Spain). After performing manual segmentation of the LA wall, the software calculated the image intensity ratio (11) and assigned each voxel as abnormal or normal LA myocardium according to the following image intensity ratio cut points (8): ≤ 1.20 defined as normal tissue, > 1.20 and ≤ 1.32 defined as interstitial fibrosis, and > 1.32 defined as dense scar. The software also calculated LA volume based on segmentation of the LA wall. The same reader repeated the analysis at least 2 weeks apart to assess intrareader repeatability in the quantification of LA fibrosis.
Statistical Analysis
The statistical analyses were conducted by one reader (S.G.) using an SPSS software package (version 23.0, IBM, Armonk, NY). The Cohen κ coefficient was calculated with previously described interpretation definitions (15) to determine interreader variability in visual scores. Assuming nonnormal distribution for the individual visual scores graded by average reader, the Mann-Whitney U test was used to detect differences between two groups. Two-tailed unpaired t test was used to detect differences in continuous variables. Intraclass correlation coefficient (ICC) with model one and type one was computed to determine intrareader agreement in the quantification of LA fibrosis. A P value < .05 was considered significant for all statistical tests.
Results
All 25 study participants completed 3D LA LGE MRI, which on average required 5.9 minutes ± 1.3 for imaging time and 63.3 minutes ± 11.0 for offline XD-GRASP reconstruction. Mean imaging time was significantly shorter (P < .01) for the proposed pulse sequence than the reference pulse sequence (10.6 minutes ± 2). Figure 3 shows comparisons of best, average, and worst examples between the proposed and reference 3D LA LGE pulse sequences. For the corresponding dynamic display of all slices, see Movies 2–7 (supplement).
Movie 2:
Video display through slice of the best quality image series obtained using the proposed pulse sequence corresponding to Figure 3 (upper left).
Movie 3:
Video display through slice of an average quality image series obtained using the proposed pulse sequence corresponding to Figure 3 (upper middle).
Movie 4:
Video display through slice of the worst quality image series obtained using the proposed pulse sequence corresponding to Figure 3 (upper right).
Movie 5:
Video display through slice of the best quality image series obtained using the reference pulse sequence corresponding to Figure 3 (lower left).
Movie 6:
Video display through slice of an average quality image series obtained using the reference pulse sequence corresponding to Figure 3 (lower middle).
Movie 7:
Video display through slice of the worst quality image series obtained using the reference pulse sequence corresponding to Figure 3 (lower right).
The average reader visual scores are summarized in Table E1 (supplement). Despite lower severity of fibrosis (ie, less hyperenhanced voxels) for the prospective cohort images obtained using the proposed pulse sequence (see below for fibrosis severity), the median conspicuity, noise, artifact, and SVS scores were all significantly higher (P < .01) for the proposed (4, 3.5, 3.5, and 11, respectively) than for the reference pulse sequence (3, 3, 3, and 9.5, respectively). For interrater reproducibility in visual scores, there was substantial to almost perfect agreement for the proposed 3D LA LGE (κ = 0.78–1) and moderate to substantial agreement for the reference 3D LA LGE (κ = 0.44–0.75).
Figure 4 shows representative fibrosis maps of patients with AF representing low and high fibrosis severity from both cohorts. Summarizing the results over 25 patients, the mean LA fibrosis was 5.6% ± 4.0 for the prospective participants and 12.2% ± 10.2 for the reference patient cohort. When separated by prior ablation history, the mean LA fibrosis in the prospective cohort was 5.3% ± 4.1 for the subgroup without prior ablation and 6.5% ± 3.7 for the subgroup who had undergone prior ablation. The mean LA fibrosis in the reference cohort was 10.4% ± 9.5 for the subgroup without prior ablation and 16.7% ± 11.3 for the subgroup who had undergone prior ablation. Given the higher severity of fibrosis (ie, more hyperenhanced voxels) for the reference cohort, intrareader repeatability in fibrosis quantification was higher for the reference cohort (ICC = 0.94) than the prospective cohort (ICC = 0.79).
Figure 4:
Example left atrial (LA) fibrosis maps of four patients with atrial fibrillation representing low (left) and high (right) severity of LA fibrosis: proposed pulse sequence (top row) and reference pulse sequence (bottom row). Top left (A): 53-year-old man; top right (B): 75-year-old woman; bottom left (C): 75-year-old woman; bottom right (D): 65-year-old woman; blue = no fibrosis; gray = interstitial fibrosis; red = dense scarring.
Discussion
This study described the development of a 3.3-fold accelerated 3D LA LGE pulse sequence using a balanced SSFP readout with stack-of-stars k-space sampling and XD-GRASP reconstruction with 100% gating efficiency. The results obtained from 25 study participants with AF imaged at 1.5 T demonstrated that the proposed 3D LA LGE sequence was capable of producing clinically acceptable image quality (median SVS = 11) with 1.5 mm × 1.5 mm × 2-mm nominal spatial resolution and 6-minute predictable imaging time.
There were several reasons as to why our proposed sequence improved on currently available 3D LA LGE pulse sequences. First, most published 3D LGE pulse sequences (16) used a gradient-echo readout. However, our experience with 3D LA LGE using a gradient-echo readout at high spatial resolution suggests inadequate SNR at 1.5 T (6). This is one of the reasons experts in 3D LA LGE favor 3-T scanners. Our approach to raising the SNR at 1.5 T centered on balanced SSFP readout and XD-GRASP reconstruction (ie, denoising). Second, our pulse sequence achieved a predictable imaging time by attaining 100% gating efficiency using the XD-GRASP framework. This was an advantage over a standard navigator-gated 3D LGE pulse sequence with navigator gating efficiency that is dependent on both the patient’s breathing pattern and technologist expertise. Third, the reference 3D LA LGE pulse sequences (6) used an acceleration factor of two. Assuming navigator gating efficiency of 40%, the relative imaging time with respect to Nyquist sampling with 100% gating efficiency would be 1.25 (ie, 25% longer than Nyquist sampling with 100% gating efficiency). For our 3.3-fold accelerated acquisition with 100% navigator efficiency, the relative imaging time with respect to Nyquist sampling was 0.30. As such, compared with a prior two-fold accelerated 3D LA LGE pulse sequence with navigator gating efficiency of 40%, our 3.3-fold accelerated 3D LA LGE with 100% gating efficiency would be 4.2 times faster, which is an improvement for lengthy 3D acquisitions, such as 3D LA LGE. Fourth, despite lower severity of fibrosis for the prospective cohort, the interreader reproducibility in visual scores was higher for the prospective cohort because the proposed pulse sequence produced higher visual scores. The intrareader repeatability in fibrosis quantification was higher for the reference cohort than the prospective cohort because it was easier for the reader to draw the left atrial contour with more hyperenhanced voxels.
This study had several limitations worth emphasizing. First, we lack an independent reference method because we did not compare our 3D LA LGE pulse sequence to a previously described 3D LGE pulse sequence (6,12) in our prospectively enrolled study participants because it would have been difficult to add another lengthy (approximately 11 minutes) research pulse sequence onto a clinical MRI examination without significantly altering the clinical workflow or avoiding gadolinium washout effects. It is also not trivial to exactly implement prior pulse sequences because there are subtle, and yet important, details not reported in publications. For these reasons, we elected to compare our results with historic images obtained at the second institution using a reference 3D LA LGE sequence (6). Second, during the image scoring process, the readers could easily discern which image came from which sequence because of the differences in image orientation. The reference 3D LA LGE sequence is performed in an axial orientation because it is the most straightforward orientation for positioning the imaging slab and navigator echo on the diaphragm outside of the slab. However, for our proposed 3D LA LGE sequence, we chose to use an oblique sagittal orientation to faithfully track diaphragm motion by including the liver in the slab and evaluate fibrosis in the left ventricle, as needed. Although we considered reformatting the DICOM files during image grading so that both the reference and proposed images were matching in orientation, we felt it was more important to have readers grade the images in native orientation. Third, in addition to the imaging planes for each sequence being different, the proposed and reference examinations used different contrast agents and doses (0.20 mmol/kg gadobutrol and 0.10 mmol/kg gadobenate dimeglumine, respectively). This decision was unavoidable because 3D LGE images were obtained at two different institutions using different gadolinium protocols. It should be noted that gadobenate dimeglumine has a higher relaxivity than gadobutrol, favoring gadobenate dimeglumine (17). Fourth, the retrospective cohort (12.2%) had a higher severity of fibrosis than the prospective cohort (5.6%), which favors repeatability in LA fibrosis quantification for the retrospective cohort. This discrepancy in fibrosis severity reflects differences in patient population (and clinical practice at two different institutions). In practice, it is difficult to exactly match clinical characteristics and fibrosis severity of two different cohorts from two different institutions. Fifth, for this initial development work, only 25 participants with AF were studied. Sixth, we did not consider the impact of arrhythmia during XD-GRASP reconstruction. Irregular heart rhythm during MRI may introduce image artifacts (blurring or ghosting) arising from misalignment of the LA wall. One approach to address this problem is to expand the XD-GRASP framework to additionally bin the data into two different heartbeat durations (ie, fifth dimension), at the expense of SNR and residual aliasing artifacts. Seventh, the mean image reconstruction time was approximately 1 hour, which is not ideal for clinical translation. A future study is warranted to explore more rapid image reconstruction pipelines, such as deep learning, and how to best take advantage of the multiple respiratory states by using motion correction either through diffeomorphism or deep learning. Eighth, this study did not include quantitative metrics, such as contrast-to-noise ratio, because of the following challenges: (a) it would be difficult to accurately contour the endocardial and epicardial LA wall and (b) noise is poorly defined and difficult to estimate in both GRAPPA (reference) and XD-GRASP (proposed) reconstructions. Last, our fibrosis measurements were not applied for clinical use. Future work includes LA fibrosis-based risk stratification systems for predicting recurrence of AF after catheter ablation in a large cohort of participants.
In summary, this study demonstrated that a 3.3-fold accelerated 3D LA LGE pulse sequence using balanced SSFP readout with stack-of-stars k-space sampling and XD-GRASP reconstruction with 100% gating efficiency was capable of producing clinically acceptable image quality with 1.5 mm × 1.5 mm × 2 mm nominal spatial resolution and 6-minute predictable imaging time in patients with AF at 1.5 T.
APPENDIX
SUPPLEMENTAL FIGURES
Disclosures of Conflicts of Interest: S.G. disclosed no relevant relationships. H.H.V. disclosed no relevant relationships. D.C.L. Activities related to the present article: institution receives NIH grant. Activities not related to the present article: disclosed no relevant relationships. Other relationships: disclosed no relevant relationships. R.J.A. disclosed no relevant relationships. B.D.W. disclosed no relevant relationships. M.I. disclosed no relevant relationships. M.M. Activities related to the present article: disclosed no relevant relationships. Activities not related to the present article: author is consultant for Circle Cardiovascular Imaging; institution receives grants from Circle Cardiovascular Imaging, Cryolife, and Siemens. Other relationships: disclosed no relevant relationships. R.S.P. Activities related to the present article: institution receives grant from American Heart Association (AHA AF SFRN). Activities not related to the present article: author is consultant for Medtronic, Abbott, and Johnson & Johnson; author is paid for expert testimony; author receives royalties from UpToDate. Other relationships: disclosed no relevant relationships. E.G.K. Activities related to the present article: disclosed no relevant relationships. Activities not related to the present article: author is consultant for Marrek; institution (University of Utah) receives money for patents; institution receives royalties from Marrek; author has stock in Marrek. Other relationships: institution receives money for issued patents and for licensed/royalties/licensee patents with Marrek. D.K. Activities related to the present article: institution received funding from NIH (R01HL116895) for projects related to this study. Activities not related to the present article: disclosed no relevant relationships. Other relationships: disclosed no relevant relationships.
Supported in part by the National Institutes of Health (grants R01HL116895, R01HL138578, R21EB024315, R21AG055954, and R01HL151079) and American Heart Association (grants 14SFRN20480260 and 19IPLOI34760317).
Abbreviations:
- AF
- atrial fibrillation
- ECG
- electrocardiography
- GRAPPA
- generalized autocalibrating partially parallel acquisition
- ICC
- intraclass correlation coefficient
- LA
- left atrium
- LGE
- late gadolinium enhancement
- SNR
- signal-to-noise ratio
- SSFP
- steady-state free precession
- SVS
- summed visual score
- 3D
- three-dimensional
- XD-GRASP
- extra motion-state golden-angle radial sparse parallel
References
- 1.Anter E, Jessup M, Callans DJ. Atrial fibrillation and heart failure: treatment considerations for a dual epidemic. Circulation 2009;119(18):2516–2525. [DOI] [PubMed] [Google Scholar]
- 2.Khan AR, Khan S, Sheikh MA, Khuder S, Grubb B, Moukarbel GV. Catheter ablation and antiarrhythmic drug therapy as first- or second-line therapy in the management of atrial fibrillation: systematic review and meta-analysis. Circ Arrhythm Electrophysiol 2014;7(5):853–860. [DOI] [PubMed] [Google Scholar]
- 3.Burstein B, Nattel S. Atrial fibrosis: mechanisms and clinical relevance in atrial fibrillation. J Am Coll Cardiol 2008;51(8):802–809. [DOI] [PubMed] [Google Scholar]
- 4.Goldberger JJ, Arora R, Green D, et al. Evaluating the Atrial Myopathy Underlying Atrial Fibrillation: Identifying the Arrhythmogenic and Thrombogenic Substrate. Circulation 2015;132(4):278–291. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Peters DC, Wylie JV, Hauser TH, et al. Detection of pulmonary vein and left atrial scar after catheter ablation with three-dimensional navigator-gated delayed enhancement MR imaging: initial experience. Radiology 2007;243(3):690–695. [DOI] [PubMed] [Google Scholar]
- 6.Marrouche NF, Wilber D, Hindricks G, et al. Association of atrial tissue fibrosis identified by delayed enhancement MRI and atrial fibrillation catheter ablation: the DECAAF study. JAMA 2014;311(5):498–506. [DOI] [PubMed] [Google Scholar]
- 7.Khurram IM, Habibi M, Gucuk Ipek E, et al. Left Atrial LGE and Arrhythmia Recurrence Following Pulmonary Vein Isolation for Paroxysmal and Persistent AF. JACC Cardiovasc Imaging 2016;9(2):142–148. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Benito EM, Carlosena-Remirez A, Guasch E, et al. Left atrial fibrosis quantification by late gadolinium-enhanced magnetic resonance: a new method to standardize the thresholds for reproducibility. Europace 2017;19(8):1272–1279. [DOI] [PubMed] [Google Scholar]
- 9.Akoum N, Daccarett M, McGann C, et al. Atrial fibrosis helps select the appropriate patient and strategy in catheter ablation of atrial fibrillation: a DE-MRI guided approach. J Cardiovasc Electrophysiol 2011;22(1):16–22. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Chubb H, Karim R, Roujol S, et al. The reproducibility of late gadolinium enhancement cardiovascular magnetic resonance imaging of post-ablation atrial scar: a cross-over study. J Cardiovasc Magn Reson 2018;20(1):21. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Khurram IM, Beinart R, Zipunnikov V, et al. Magnetic resonance image intensity ratio, a normalized measure to enable interpatient comparability of left atrial fibrosis. Heart Rhythm 2014;11(1):85–92. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Pontecorboli G, Figueras I Ventura RM, Carlosena A, et al. Use of delayed-enhancement magnetic resonance imaging for fibrosis detection in the atria: a review. Europace 2017;19(2):180–189. [DOI] [PubMed] [Google Scholar]
- 13.Feng L, Axel L, Chandarana H, Block KT, Sodickson DK, Otazo R. XD-GRASP: Golden-angle radial MRI with reconstruction of extra motion-state dimensions using compressed sensing. Magn Reson Med 2016;75(2):775–788. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Griswold MA, Jakob PM, Heidemann RM, et al. Generalized autocalibrating partially parallel acquisitions (GRAPPA). Magn Reson Med 2002;47(6):1202–1210. [DOI] [PubMed] [Google Scholar]
- 15.Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics 1977;33(1):159–174. [PubMed] [Google Scholar]
- 16.Foo TK, Stanley DW, Castillo E, et al. Myocardial viability: breath-hold 3D MR imaging of delayed hyperenhancement with variable sampling in time. Radiology 2004;230(3):845–851. [DOI] [PubMed] [Google Scholar]
- 17.Rahsepar AA, Ghasemiesfe A, Suwa K, et al. Comprehensive evaluation of macroscopic and microscopic myocardial fibrosis by cardiac MR: intra-individual comparison of gadobutrol versus gadoterate meglumine. Eur Radiol 2019;29(8):4357–4367. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.