0784

Cardiac MR fingerprinting with a short acquisition window in healthy volunteers and 62 consecutive patients referred for clinical CMR

Simone Rumac¹, Anna Giulia Pavon², Jesse Hamilton³, David Rodrigues¹, Nicole Seiberlich³, Juerg Schwitter², and Ruud B. van Heeswijk¹
¹Department of Radiology, Lausanne University Hospital (CHUV), Lausanne, Switzerland, ²Cardiology Service, Lausanne University Hospital (CHUV), Lausanne, Switzerland, ³Department of Radiology, University of Michigan, Ann Arbor, MI, United States

Synopsis

Cardiac magnetic resonance fingerprinting (cMRF) can be used to simultaneously acquire myocardial T₁ and T₂ maps in a single breath-hold. However, the common 250 ms acquisition window of cMRF might leave it vulnerable to motion artefacts. The goal of this study was therefore to compare the performance of cMRF with a short acquisition window (150ms) and low-rank reconstruction to that of routine cardiac parametric mapping techniques. In 7 healthy volunteers, and 62 cardiac patients, cMRF resulted in similar native relaxation times, but slightly different post-contrast T₁ and ECV values compared to routine techniques.

Introduction

Cardiac magnetic resonance fingerprinting (cMRF[1]) has been demonstrated to be a robust and accurate T₁ and T₂ mapping technique. However, the common 250ms acquisition window of cMRF might leave it vulnerable to motion artifacts in patients with high heart rates. A shorter acquisition window would thus be desirable, especially if the loss in precision due to less acquired signal can be compensated with a low-rank reconstruction. The goal of this study was therefore to compare the accuracy and robustness of a cMRF sequence with a short acquisition window and low-rank reconstruction[2] to routine T₁, T₂, and ECV mapping techniques. The comparison was performed both in a small cohort of healthy volunteers and in a heterogeneous group of consecutive patients referred for clinical CMR.

Methods

The accuracy of cMRF with a short acquisition window was first compared to reference and clinical routine parameter mapping techniques (MOLLI[3]; T₂-prepared bSSFP[4]) in the ISMRM-NIST phantom (QalibreMD) at 1.5T (Sola, Siemens). Reference T₁ and T₂ relaxation times were obtained with inversion-recovery TSE and multi-echo SE, respectively. Due to the sensitivity of the routine cardiac T₂ mapping sequence to short T₁ values, only those spheres with T₁>500ms were taken into the T₂ analysis. In vivo, cMRF was performed with the following parameters in both subject groups: 29 readouts/heartbeat, duration 15 heartbeats, pixel size=1.6x1.6mm2, slice thickness=8mm3, acquisition window=150ms. For each slice, a heart-rate dependent low-rank dictionary was created and used to reconstruct the parametric maps. The dictionaries were designed to take the slice excitation profile and B₁⁺ inhomogeneity into account [2]. The human study was approved by the Institutional Review Board, and participants provided informed consent. In the healthy volunteers group (n=7, average age=27, 80% female), routine T₁ and T₂ maps and cMRF were acquired at three short-axis and a four-chamber orientation. In consecutive patients referred for CMR (n=62, average age=58y, 30% female), routine native T₁ (n=62) and T₂ (n=12) mapping as well as cMRF (n=62) were acquired in one basal slice; routine T₁ mapping (n=47) and cMRF (n=47) were also performed 20-25 minutes after gadolinium injection (0.2 ml/kg Dotarem). The parameter maps were manually segmented, and the synthetic ECV was calculated[6]. Linear regression against the reference values was used to assess accuracy in the phantom, while Student’s t-tests and Bland-Altman analyses were used to assess differences between routine techniques and short acquisition window cMRF in vivo.

Results

The phantom mapping demonstrated similar or higher T₁ and T₂ accuracy of the cMRF over a wider range than the routine mapping techniques (Fig.1). In the healthy volunteers (Table 1), the cMRF myocardial T₁ and T₂ values showed small but non-significant differences compared to MOLLI (cMRF: 1019±90ms; Routine: 1001±48ms, p=0.28) and T₂-prepared bSSFP (cMRF: 43±4ms; Routine: 46±4ms, p=0.02). In the patients, both the native T₁ (cMRF: 1011±61ms; Routine: 1028±56ms, P=0.17) and T₂ (cMRF: 44±7ms; Routine: 46±3ms, p=0.53) values confirmed the good agreement (Fig.2). However, post-contrast myocardial T₁ values (Fig.2C-D, Fig.3B) were lower than the routine values (cMRF: 391±43ms; Routine: 441±43ms, p<0.001), while the blood pool values did not differ (cMRF: 268±42ms; Routine: 282±47ms, p=0.23). This was then reflected by slightly higher estimations of the synthetic ECV (cMRF: 28±4%; Routine: 26±3%; p=0.02).

Discussion

cMRF with a short acquisition window and low-rank reconstruction performed similarly or better than routine techniques when tested against reference relaxation times obtained from the NIST phantom. In vivo, the overall average cMRF and routine relaxation times appeared to be highly similar in healthy volunteers and patients. The small but significant T₂ difference observed in healthy volunteers might be due to the small sample size. The difference in post-contrast myocardial T₁ is more significant, and might for example be caused by partial volume effects, through-plane motion, or a T₂ influence on the 4(1)3(1)2 MOLLI fit[7]. These findings warrant that ECV be measured with cMRF in a cohort of healthy volunteers in order to establish healthy reference values. Overall, we conclude that in 62 consecutive patients, cMRF with a short acquisition window and low-rank reconstruction resulted in comparable native cardiac T₁ and T₂ values when compared to routine techniques, but resulted in slightly different post-contrast myocardial T₁ and ECV estimations.

Acknowledgements

Swiss Heart Foundation and Swiss National Science Foundation (32003B_182615)

References

1. Y. Liu et al., JACC: Cardiovascular Imaging. 2018; 11(12), 1837–1853

2. J. I. Hamilton et al., Magn. Reson. Imaging. 2018; 53(May), 40–51

3. D. R. Messroghli et al., Magn Reson Med. 2004, 52(1):141-6

4. T. Y. Huang et al., Magn. Reson. Imaging. 2007; 57(5), 960–966

5. M. D. Cerqueira et al., Circulation. 2002; 105(4), 539–542

6. T.A. Treibel et al., JACC: Cardiovascular Imaging. 2016; 9(1), 54–63

7. P. Kellman et al., J. Cardiovasc. Magn. R. 2014; 16(1), 1–20.

Figures

Figure 1. Accuracy of cMRF compared to reference techniques in the ISMRM-NIST phantom. A) T₁ map of the reference phantom obtained with short acquisition window cMRF. B) Linear regression plots of T₁ values in the phantom in the post-contrast myocardial T₁ range. C) Linear regression plots of T₁ values in the phantom in the native myocardial T₁ mapping range. cMRF is as accurate at the long and short T₁ values as the dedicated MOLLI sequences are. D) Linear regression plots for the T₂ values. The lower slopes are likely due to the contributions of stimulated echoes to the multi-echo SE.

Figure 2. Representative T₁ and T₂ maps acquired with cMRF (top row) and routine technique in several patients. A,B) T₁ maps before contrast agent injection (72 y.o., male). The maps have comparable precision and accuracy. A small artefact from a stent can be observed at the posterior right ventricle insertion point in both maps. C,D) T₁ maps 20-25 minutes after contrast agent injection (27 y.o., male). The T₁ values are slightly higher in the centre of the myocardium of the routine maps. E,F) Native T₂ maps (57 y.o., female).

Figure 3. Bland-Altman analyses of cMRF and routine techniques for the entire myocardium in patients. A) Native T₁ relaxation times obtained with cMRF vs. 5(3)3 MOLLI. B) Post-contrast T₁ relaxation times obtained with cMRF vs. 4(1)3(1)2 MOLLI. A significant bias was observed, while the post-contrast LV blood T₁ was not significantly different between the two techniques. C) Native T₂ relaxation times obtained with cMRF vs. T₂-prepared bSSFP. D) ECV calculated with cMRF vs. the routine techniques.

Table 1. The average myocardial relaxation times and ECV measured with cMRF and routine techniques. Similar native relaxation times were measured with all techniques and in both the healthy volunteers and patients. However, the post-contrast myocardial T₁ values differed between cMRF and MOLLI, which together with similar post-contrast blood T₁ values resulted in slightly different ECV estimations.

Proc. Intl. Soc. Mag. Reson. Med. 28 (2020)

0784