Synopsis

Keywords: Heart, Oxygenation

A breath-holding non-cardiac gated 3D stack-of-spiral data acquisition scheme was developed to continuously sample the data for cardiac quantitative susceptibility mapping. Compared to the previously proposed navigator-based prospective Cartesian acquisition, the accelerated spiral sequence can be done within 20 seconds breath-holds, leading to over a 20-fold reduction in scan time. The spiral QSM as well as the navigator QSM were performed on cohorts of healthy volunteers and COVID-19 survivors, showing well aligned quantification results on the differential blood oxygenation between the right and left heart.

Introduction

Cardiac Quantitative Susceptibility Mapping (QSM)¹ has been developed to non-invasively measure the differential blood oxygen saturation (ΔSaO₂) between the right and left heart. The novel cardiac magnetic resonance imaging technique utilizes the susceptibility contrast derived from deoxygenated blood pool. Previously, a prospective data sampling strategy based on free-breathing 1D diaphragmatic navigator and ECG triggered 3D Cartesian multi-echo gradient echo acquisition was proposed to acquire the multi-echo gradient echo data needed for QSM². However, this method requires a complicated scanning setup and suffers from low scanning efficiency, depending on the subject’s respiratory and cardiac movements. To speed up the acquisition, a breath-holding non-cardiac gated 3D stack-of-spiral scheme is used in this study to continuously sample the data for cardiac QSM by exploiting the high efficiency and motion robustness of non-Cartesian acquisition, and the ΔSaO₂ quantification is compared with the navigator method.

Methods

The accelerated cardiac QSM data acquisition used a 3D spoiled multi-echo stack-of-spiral sequence. A variable density spiral trajectory was designed to have a 2-fold oversampling ratio in the center of k-space with the edges undersampled by the factor of 0.7, to achieve a fully sampled k-space with 36 leaves, based on the given gradient strength and slew rate limit.³ The multi-echo signal was sampled along the same spiral leaf in each repetition time and across all slice encodes before moving on to the consecutive golden-angle rotated leaf. Scan time was 20sec using a breath-hold covering the whole heart in the axial plane with the following imaging parameters: number of spiral leaves N_l = 36, each spiral leaf readout points N_sp = 1096, readout bandwidth ±125kHz, flip angle FA = 12º, number of echoes N_e = 3, TE₁/TR/ΔTE = 0.4/17.8/5.9msec, reconstructed matrix size 256×256×24, image resolution 1.8×1.8×5mm³. A reference free induction decay (FID) signal at each kz encoding was acquired at the beginning of the scan for off resonance correction.⁴ The spiral QSM sequence was performed on a 3T scanner (GE750) on healthy volunteers as well as on a 3T clinical scanner (GE PET/MR) on COVID-19 survivors. For comparison, Cartesian navigator QSM (“NAV”) was acquired at the same time on all the subjects on axial plane, with the following parameters: readout bandwidth ±83.33kHz, flip angle FA = 15º, number of echoes N_e = 5, TE₁/TR/ΔTE = 1.4/18.7/3.4msec, reconstructed matrix size 256×256×24, image resolution 1.8×1.8×5mm³. A GRAPPA factor of 2 was implemented in NAV QSM to reduce scan time. All subjects provided consent for this IRB approved protocol.

The acquired spiral raw k-space phase data were first demodulated along the readout direction by the estimated field from reference FID on a slice-by-slice fashion for the purpose of off resonance correction. Gridding was applied on the fully sampled spiral to reconstruct the multi-echo complex images.³ Next, the field map was fitted from the multi-echo complex data and unwrapped by graph-cut based method⁵, with iterative decomposition of water and fat with echo asymmetry and least squares estimation (IDEAL)⁶ for water fat separation. Then total field inversion (TFI+0)^2,7,8 together with regularization of blood pool susceptibility variation was performed:
$$y^*=arg\min_y ||w(f-DPy)||^2_2 + \lambda ||M_G\triangledown Py||_1 + \sum_i \lambda_i ||M_iP(y-\overline{y}^i)||^2_2$$
The first term imposes data fidelity on the dipole field convolution; the second and third term introduce l-1 regularization and additional blood pool uniformity regularization inside each heart chamber region, segmented from the combined-echo magnitude, respectively. ΔSaO₂ is estimated from the difference between the mean susceptibility in right/left ventricle blood pools scaled by hematocrit. ΔSaO₂ estimated from spiral QSM were compared with the navigator QSM by Deming regression and Bland Altman plots. Image reconstruction was performed using in-house C++ program. QSM reconstruction were conducted on MATLAB R2020b. Statistical analysis was performed using R 4.2.1.

Results

9 healthy volunteers as well as 14 COVID-19 survivors (age: 44±19, 52% male, N=23) underwent both spiral QSM and Cartesian NAV cardiac QSM all successfully. The Cartesian NAV acquisition time was 412±110sec, with navigator efficiency 37.7±8.8%. Spiral QSM reduced it within 20sec, over a factor of 20. Among all the subjects, the ΔSaO₂ from spiral was 16.84±4.53%, compared to 16.43±4.70% for reference NAV acquisition.

Figure 1 shows a representative case with the reconstructed QSM from reference NAV and spiral acquisition. Method comparison analysis (Figure 2) shows that spiral QSM has consistent quantification on ΔSaO₂, compared with Cartesian NAV (ΔSaO_2,sp = 0.96ΔSaO_2,nav + 1.23 (%), Pearson’s r = 0.81).

Discussion

The proposed non-gated spiral cardiac QSM accelerated the scan within 20sec breath-holds, while achieving a consistent ΔSaO₂ quantification compared to Cartesian NAV method. Validation of the proposed spiral QSM by comparing with right heart catheterization (RHC) is needed. Figure 3 shows such a comparison in one pulmonary hypertension patient (ΔSaO₂ value: 28% RHC vs 22.5% NAV vs 26.8% spiral). Future work will introduce retrospective motion compensation/correction to overcome motion-induced blurring.

Conclusion

The accelerated stack-of-spiral cardiac QSM reduced the scan time over a factor of 20 compared to the prospective navigator method with consistent ΔSaO₂ quantification.

Acknowledgements

No acknowledgement found.