2181

Realistic Simulation of High-Performance Low Field Cardiac Cine Imaging

Xitong Wang¹, Namgyun Lee², Adrienne E. Campbell-Washburn³, and Krishna S. Nayak¹
¹Electrical Engineering, University of Southern California, Los Angeles, CA, United States, ²Biomedical Engineering, University of Southern California, Los Angeles, CA, United States, ³National Heart, Lung, and Blood Insititute, National Institues of Health, Bethesda, MD, United States

Synopsis

There is renewed interest in high-performance low field imaging, especially for applications that are limited by susceptibility and SAR, such as cardiac CINE bSSFP imaging. Here, we demonstrate a framework for realistic simulation of 3D CINE bSSFP cardiac imaging at various field strengths. This is benchmarked against in-vivo data acquired on a high-performance 0.55 Tesla MRI. We utilize the XCAT phantom, 3D stack of spiral sampling, and realistic noise. This setup can be used to predict minimum field strength requirements and to test data sampling and reconstruction techniques.

Introduction

Over the last four decades, MRI has trended towards higher field strengths, motivated primarily by neuro, spine, and musculoskeletal applications. However, many indications benefit from lower field due to reduced SAR and susceptibility. There is renewed interest in high performance low field (HPLF) systems for advanced imaging^1,2,3,4. We must rely on simulations to explore performance as a function of B₀, because B₀ is difficult and expensive to change in real instruments.
Here, we demonstrate a framework for realistic simulation of 3D CINE bSSFP cardiac imaging at various field strengths. We utilize the XCAT phantom^6,7, 3D stack of spiral sampling, and realistic noise. This setup may be used to predict minimum field strength requirements and to test data sampling and reconstruction techniques.

Methods

Three-dimensional dynamic anatomic masks were generated using XCAT software^6,7. Simulated object parameters: 1 mm isotropic resolution; heart rate = 60 beats per minute; no respiration; and 40 RR intervals. Imaging parameters: Stack-of-Spiral bSSFP; TR/TE = 10ms/2ms; T_read = 6ms (keep constant for different field strengths); FA = optimized for maximum blood-myocardium contrast at each field strength; FOV 36x36x8cm³; resolution 2.81x2.81x4mm³, 8-arm uniform density spiral. Reconstruction: NUFFT⁹ in the k_x,k_y plane; FFT along k_z; 25 cardiac phases without view sharing. Table 1 contains the T₁, T₂ and proton density values ^5,6. Figure 1 illustrates the pipeline.
Per-sample noise was measured on a 0.55T high performance low field scanner in a healthy adult subject⁸. This HPLF system is a customized commercial 1.5T system (MAGNETOM Aera, Siemens Healthcare, Erlangen, Germany) that was ramped down to operate as a prototype 0.55T system with retained gradient performance (45 mT/m amplitude, 200 T/m/s slew rate) and a retuned cardiac coil (18-channel, 6 anterior, 12 posterior). Virtual single-channel data was generated by applying coil compression¹⁰on 18-channel k-space data for signal strength measurements (k-space origin) and per-sample noise measurements. We accounted for differences in slab thickness between the measurements (2D, 8-mm thick) and simulation (3D 80-mm thick) by assuming a 10-fold difference in signal at the k-space origin. Signal strength was scaled using the PD values and given to the Bloch simulator (see Figure 1),

P D_{s i m} = P D (B_{0} / 0.55) α

$PD_{sim} = PD (B_{0}/0.55) \alpha$ , where

B_{0}

$B_{0}$ is the simulated field strength, and

α

$\alpha$ is a scale factor determined based on the 0.55T measurements. Complex-valued Gaussian noise was generated using the measured per-sample noise standard deviation.

Results

Figure 2 compares simulated 0.55T 3D CINE images against measured 0.55T 2D CINE images (66-shot spiral out; TR/TE = 8ms/4ms;18-channel receiver). There are substantial differences in spatial resolution, slab coverage, and precise readout scheme. However, the overall quality is comparable. This includes spiral artifacts in the short-axis plane.
Figure 3 contains predictions of imaging quality for field strengths of 0.2T, 0.35, 0.55T, and 0.75T. All imaging and reconstruction parameters were the same except for bSSFP flip angles (80° for 0.2T, 76° for 0.35T, 72° for 0.55T, and 69° for 0.75T), which were optimized for maximum blood-myocardium contrast (see Figure 4). Left ventricular function is clearly visualized at 0.35T and above.

Discussion

Due to long T₂* and reduced banding artifacts, HPLF is expected to open opportunities for novel data sampling and reconstruction with longer readouts and undersampling. This simulation platform would be useful for testing these new approaches.
This framework could be used to determine minimum field strength requirements for cardiac CINE bSSFP imaging, as has been done for other applications¹¹. This would require applying standard post-processing pipelines such as LV segmentation, for computation of volumes and regional wall motion.
This work has several limitations: 1) We assume on-resonance at all spatial positions. However, off-resonance is less of an issue at low-field because of reduced off-resonance and a fixed bSSFP null-to-null spacing (with a fixed TR). This also may be less of an issue if we switch to spiral in-out or trajectories where the center of k-space is acquired at TR/2 which experiences an echo¹². 2) The XCAT simulation was scaled using in-vivo measurements from a different human subject. The object distribution impacts signal scaling, so it is a rough approximation to assume comparable signal per unit slab thickness. 3) We applied extreme coil compression (18 to 1), well beyond the intended use of this technique¹⁰. This is temporary until we develop models for coil sensivities as a function of field strength that can be applied to the XCAT phantom.

Conclusion

We have developed a simulation framework based on XCAT that is capable of predicting 3D bSSFP CINE imaging performance at various field strengths around 0.5 Tesla. Initial results indicate that single breath-hold 3D cardiac CINE imaging is viable on HPLF MRI systems. Adequate spatio-temporal resolution is demonstrated without any undersampling using only the increased efficiency of long-TR bSSFP and stack-of-spiral acquisitions.

Acknowledgements

Grant Support: NIH R01-HL130494, American Heart Association; Research Support: Siemens Healthineers; We acknowledge the assistance of Siemens Healthcare in the modification of the MRI system for operation at 0.55T under an existing cooperative research agreement between NHLBI and Siemens Healthcare.

References

1. Campbell-Washburn AE, Ramasawmy R, Restivo MC, Bhattacharya I, Basar B, Herzka DA, et al. Opportunities in Interventional and Diagnostic Imaging by Using High-performance Low-Field-Strength MRI. Radiology. 2019 Oct 1;190452.

2. Marques JP, Simonis FFJ, Webb AG. Low-field MRI: An MR physics perspective. J Magn Reson Imaging. 2019 Jan 13;58:1182–15.

3. Simonetti OP, Ahmad R. Low-Field Cardiac Magnetic Resonance Imaging: A Compelling Case for Cardiac Magnetic Resonance's Future. Circ Cardiovasc Imaging. 2017 Jun;10(6).

4. Stainsby JA, et al. Imaging at 0.5 T with high-performance system components. Proc. ISMRM 27th Scientific Session, Montreal, May 2019. p1194.

5. Bottomley, P. A., Foster, T. H., Argersinger, R. E. and Pfeifer, L. M. (1984), A review of normal tissue hydrogen NMR relaxation times and relaxation mechanisms from 1–100 MHz: Dependence on tissue type, NMR frequency, temperature, species, excision, and age. Med. Phys., 11: 425-448.

6. Segars, W. P., Sturgeon, G., Mendonca, S., Grimes, J., & Tsui, B. M. (2010). 4D XCAT phantom for multimodality imaging research. Medical physics, 37(9), 4902–4915.

7. Wissmann, L., Santelli, C., Segars, W. P., & Kozerke, S. (2014). MRXCAT: Realistic numerical phantoms for cardiovascular magnetic resonance. Journal of cardiovascular magnetic resonance, 16(1), 63.

8. Restivo M, et al. Long TR bSSFP Cardiac Cine Imaging at Low Field (0.55T) using EPI and Spiral Sequences for Improved Sampling Efficiency. Proc. ISMRM 27th Scientific Session, Montreal, May 2019. p1084.

9. Fessler JA. Image reconstruction toolbox. [http://web.eecs.umich.edu/~fessler/code/index.html]

10. Zhang T, Pauly JM, Vasanawala SS, Lustig M. "Coil Compression for Accelerated Imaging with Cartesian Sampling," Magn Reson Med 2013; 69: 571-58211. Wu Z, Chen W, Nayak KS. Minimum Field Strength Simulator for Proton Density Weighted MRI. PLoS ONE. 2016;11(5):e0154711.

11. Wu Z, Chen W, Nayak KS. Minimum Field Strength Simulator for Proton Density Weighted MRI. PLoS ONE. 2016;11(5):e0154711.

12. Scheffler K, Hennig J. Is TrueFISP a gradient-echo or a spin-echo sequence? Magn Reson Med. 2003 Jan 31;49(2):395–7.

Figures

TABLE 1– Simulated tissue parameters. ^† from Segars et al. ¥ from Campbell et al. ^ƒ based on Bottomley et al. but adjusted to match Campbell et al. at 0.55T. * parameters assumed to be independent of field strength.

FIGURE 1 – Simulation pipeline. Dynamic XCAT anatomical masks are generated first. Anatomical masks, T₁(dependent on field strength), T₂, and PD (proton density)are input to a Bloch simulator of the steady-state bSSFP signal. Non-Cartesian encoding is simulated by applying NUFFT along x,y and FFT along z. Field strength-dependent noise (i.i.d. bivariate Gaussian) is then added to k-space data. Finally, image reconstruction is performed using iNUFFT along k_x,k_y and iFFT along k_z.

FIGURE 2 – Visual comparison of experiment and simulated short-axis bSSFP CINE images at 0.55T. Still frames are shown for (left) mid-diastole and (right) end-systole.

FIGURE 3 – Simulation of 3D bSSFP CINE image quality for a range of B₀ field strength (0.2T, 0.35T, 0.55, and 0.75T). Shown are mid-short axis, two-chamber, and four-chamber sections of the 3D dynamic datasets.

FIGURE 4 – Simulated flip angle as a function of B0 field strength. We use the flip angle that maximizes on-resonance blood-myocardium contrast as predicted by Bloch simulation for bSSFP imaging. This flip angle selection criteria is resonable because CINE bSSFP is primarily used for the evaluation of left ventricular function, wall motion, and regional thickening, which requires clear identification of the blood-myocardium boundary.

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

2181