3244

Quantitative Evaluation of Self-Gating and nonuniform self-gating for highly irregular respiratory patterns
Patrick Metze1, Tobias Speidel1, Fabian Straubmüller1, and Volker Rasche1,2
1Internal Medicine II, Ulm University Medical Center, Ulm, Germany, 2Core Facility Small Animal Imaging (CF-SANI), Ulm University, Ulm, Germany

Synopsis

In this work we evaluate the proposed nonuniform self-gating (nuSG) method for lung imaging in highly irregular respiratory patterns and compare its performance to image-based and k-space-based gating methods and to a breath-hold reference. Functional parameters (proton fraction, fractional ventilation) and image sharpness are evaluated, with the nuSG algorithm showing a superior performance in case of fractional ventilation and image sharpness. In more uniform motion patterns its performance is similar to image-based SG and both techniques outperform k-space based gating.

Introduction/Purpose:

MRI lung imaging has emerged as a promising alternative to computed tomography due to its capabilities in the visualization and assessment of functional and morphological information, without the use of ionizing radiation1,2. However, short T2* relaxation times and low proton densities in the lung lead to low SNR, with scan durations being limited by maximum breath-hold durations. Dedicated sequences such as ultrashort echo time (UTE3) and zero echo time (ZTE4) better exploit the available signal and have furthermore been optimized for lung imaging5-9. While breath-hold imaging is possible for healthy and cooperative subjects10,11,12, patients with lung diseases often exhibit limited breath-hold capabilities, leading to the need of free-breathing acquisitions in combination with motion compensation strategies. Traditional self-gating approaches rely on the identification of a characteristic frequency and thus perform best for uniform motion, but often irregular breathing patterns due to anxiety, shortness of breath or coughing are observed. Therefore, we adapted the nonuniform self-gating (nuSG) framework13,14,15 for lung imaging and evaluated the method in six healthy volunteers, comparing the results to established k-space16 and image-based9,17 gating techniques.

Methods:

All imaging experiments were performed on a clinical 1.5 T whole-body MRI system (Achieva 1.5 T, Philips Healthcare, Best, The Netherlands) using a dedicated 32-channel cardiac coil. A 2D UTE sequence (parameters are given in Table 1) was combined with a tiny golden angle (tyGA) acquisition scheme18,19 to minimize motion sensitivity, eddy current related artefacts and to provide a better angular coverage for sliding-window and self-gating reconstructions. At two coronal slice locations two breath-hold acquisitions (inspiration and expiration) and two free-breathing acquisitions with uniform and irregular respiratory patterns were performed. The free-breathing acquisitions were reconstructed using nuSG13,14,15, image-based self-gating9 and a k-space center based self-gating16 (Figure 1). To obtain quantitative information on lung density, proton fraction values (PF) in inspiration and expiration were pixelwise calculated according to $$PF = \frac{SI_{lung}}{SI_{muscle}}\textrm{exp}\left(\frac{TE}{T^*_2}\right),$$ with SI being the signal intensity of the lung parenchyma and intercostal muscle, TE the echo time and $$$T^*_2$$$ = 2.11 ms20. After nonrigid image registration of exhaled and inhaled motion states using the Medical Image Registration Toolbox for Matlab (MIRT)21, fractional ventilation maps (FV) were calculated according to$$FV = \frac{SI_{EX}-SI_{IN}}{SI_{EX}}$$ To evaluate the residual motion blur after gating, image sharpness was analysed based on a standard metric (e.g. 13, 22) that identifies the position of 25% and 75% of the maximum signal intensity on an interpolated intensity profile placed over the lung liver interface. For image sharpness analyses, the lung-liver interface is divided into three areas that are automatically detected by slope and variance analyses. The areas are described by linear equations and the x-distance of the intersections is chosen as a measure for sharpness.

Results:

Figure 2 shows reconstructed coronal slices of all three gating approaches. For inspiration, image sharpness of the nuSG approach is clearly superior to img-based and k-space-based self-gating. Further nuSG allows the reconstruction of the breathing motion over the entire acquisition period (Figure 3). All quantitative measurements are summarized in Figure 4. Both sharpness measures show the superiority of nuSG and img-based self-gating. Sharpness for breath-hold images is higher in expiration compared to inspiration. nuSG sharpness is on a similar level across both motion states and both respiratory patterns, with proton fractions being overestimated. Differences between inspiration and expiration are noticeable for breath-holds, image-based gating and to a lesser extent in nuSG and k-space based SG. Concerning fractional ventilation, nuSG comes closest to the BH reference, with a slight underestimation.

Discussion:

The presented results are in accordance with previous studies (9, 17) and indicate that k-space based self-gating is clearly outperformed by image-based and non-uniform self-gating. Especially for irregular respiration with varying ex- and inspiratory positions, the lung liver interface appears blurred in k-space based-SG. The reduced inspiratory image sharpness for the breathhold approach is most likely caused by respiratory drifts. For uniform breathing patterns the image-based approach performed excellent but partly fails for irregular breathing. Non-uniform self-gating works equally well for data acquired during uniform and irregular breathing, thus exhibiting better sharpness measures for inspiration even compared to the BH reference. The higher proton fraction values in the nuSG reconstructions could be due to the reconstruction with less data. The resulting lower SNR increases the influence of noise and could lead to an increased lung signal in magnitude images. However, there seems to be no negative effect on FV. As the difference between inspiration and expiration is typically higher in BH imaging compared to free breathing, it is obvious that the calculated FV is higher compared to gated imaging. While nuSG and image-based SG yield similar FV, ksp-based SG yields reduced values, most likely due to residual image blur.

Conclusion:

The nuSG approach in combination with 2D-tyUTE achieves functional values in line with the image-based SG approach as well as with the reference BH technique, while offering the advantage of intrinsically reconstructing a movie of the underlying time course of the motion, yielding a possible time-resolved evaluation of lung parenchyma changes even for highly non-uniform motion.

Acknowledgements

The authors thank the Ulm University Centre for Translational Imaging MoMAN for its support.

References

1. Hirsch et al., Pediatr Radiol 2020

2. Serra et al., Chest 2011, 140 (6)

3. Holmes et al., Radiography 2005, 11 (3)

4. Weiger et al., Magn Reson Med 2013, 70 (2)

5. Bae et al., Eur Radiol 2019, 29 (5)

6. Bae et al., Eur Radiol 2020

7. Burris et al., Radiol 2016, 278 (1)

8. Zucker et al., J Magn Reson Imaging 2018, 47 (1)

9. Tibiletti et al., Magn Reson Med 2016, 75 (3)

10. Wild et al., Insights into Imaging 2012, 3 (4)

11. Biederer et al., Insights into Imaging 2012, 3 (4)

12. Lederlin et al., J Magn Reson Imaging 2014, 40 (4)

13. Wundrak et al., Magn Reson Med 2016, 76 (3)

14. Metze et al., Proceed ISMRM 2020 #0438

15. Speidel et al., Proceed ISMRM 2020 #1087

16. Weick et al., J Magn Reson Imaging 2013, 37 (3)

17. Balasch et al., J Magn Reson Imaging 2020

18. Wundrak et al., IEEE T Med Imaging, 2014, 34 (6)

19. Wundrak et al., Magn Reson Med, 2016, 75 (6)

20. Yu et al., Magn Reson Med, 2011, 66 (1)

21. Myronenko, Mirt, 2010

22. Larson et al. Magn Reson Med, 2005, 53 (1)

Figures

Figure 1: Overview of the different gating strategies. nuSG and image-based self-gating use the same high-temporal, low-spatial resolution sliding window images to calculate pairwise correlation and the navigator signal, respectively. For k-space based gating, relevant coils are pre-selected and their DC-k-space signal bandpass filtered. The resulting signal is then combined via PCA.

Table 1: Scan parameters for all acquisitions..

Figure 2: Exemplary image quality and the corresponding intensity profiles over the lung-liver interface for all three reconstruction techniques and the same acquisition. In expiration (top row) there is not substantial motion blur of the lung-liver interface, although the k-space gated image appears somewhat washed out. For inspiration, the nuSG reconstruction shows highest motion fidelity, followed by the image-gated reconstruction. The lung-liver interface is clearly blurred in the ksp-gated reconstruction.

Figure 3: Exemplary nuSG reconstructed image (right) and time course for uniform and non-uniform motion along the red intensity profile. The displacement of the lung-liver interface exhibits clearly non-uniform motion for the lower row (e.g. around frame 200 and between frame 500 and 600).

Figure 4: Quantitative evaluation of all acquisitions. Both sharpness measures (top row) show superior performance of nuSG and img-based gating compared to ksp-based gating. For inspiration, sharpness is even better in nuSG reconstruction compared to the BH reference. The proton fraction calculated for nuSG images is higher than for BHs or the other gating techniques. However, fractional ventilation calculated from nuSG comes closest to the BH reference.

Proc. Intl. Soc. Mag. Reson. Med. 29 (2021)
3244