Magnetic Resonance in Medicine 73:692–696 (2015)
A New Framework for Interleaved Scanning in Cardiovascular MR: Application to Image-Based Respiratory Motion Correction in Coronary MR Angiography Markus Henningsson,1,2,3,4* Giel Mens,5 Peter Koken,6 Jouke Smink,5 and Rene M. Botnar1,2,3,4 INTRODUCTION
Purpose: To describe a new framework for interleaving scans and demonstrate its usefulness for image-based respiratory motion correction in whole heart coronary MR angiography (CMRA). Methods: Scan interleaving using the proposed approach was achieved by switching between separately defined, independent scans at arbitrary time points during their execution, using a generic function call. The scan interleaving framework was used to perform scan interleaving for image-based respiratory navigation of CMRA with spiral, radial, and Cartesian echoplanar imaging (EPI) navigator k-space trajectories. Eight healthy volunteers were scanned. Results: Improved coronary vessel sharpness and visual scores were obtained using spiral and Cartesian EPI navigators compared with radial navigators. Conclusion: The usefulness of the proposed scan interleaving framework was demonstrated for image-based respiratory motion correction. It facilitated more direct comparisons of image navigator acquisitions with different k-space trajectories. Furthermore, we could demonstrate that spiral and Cartesian EPI navigators may be particularly suitable for image-based motion correction, as they provide improved motion correction and high navigator apparent signal-to-noise ratio while spending very little magnetization, thereby minimizing saturation C 2014 Wiley effects. Magn Reson Med 73:692–696, 2015. V Periodicals, Inc. Key words: interleaved scanning; cardiovascular MRI; respiratory motion correction
1 Division of Imaging Sciences & Biomedical Engineering, King’s College London, London, UK. 2 Wellcome Trust and EPSRC Medical Engineering Center, King’s College London, London, UK. 3 BHF Centre of Excellence, King’s College London, London, UK. 4 NIHR Biomedical Research Centre, King’s College London, London, UK.. 5 Philips Healthcare, Best, The Netherlands. 6 Philips Research, Hamburg, Germany
*Correspondence to: Markus Henningsson, Ph.D., King’s College London, Division of Imaging Sciences and Biomedical Engineering, The Rayne Institute, 4th Floor, Lambeth Wing, St. Thomas’ Hospital, London, SE1 7EH, UK. E-mail: [email protected] Correction added after online publication 24 March 2014. Author corrected the coauthor affiliation numbering and the e-mail address of the corresponding author. Received 11 June 2013; revised 6 January 2014; accepted 7 January 2014 DOI 10.1002/mrm.25149 Published online 17 March 2014 in Wiley Online Library (wileyonlinelibrary. com). C 2014 Wiley Periodicals, Inc. V
For a long time there has been a need to interleave concurrently performed scans with more or less independent imaging parameters in cardiovascular MR, including bolus tracking in contrast-enhanced MR angiography (MRA) (1), phase-sensitive inversion recovery late gadolinium enhancement (2), or acceleration techniques for time-resolved scans such as k-t BLAST (3) that uses interleaved training data sets. Perhaps the best example of interleaved scanning in cardiovascular MR is in free-breathing navigator-gated cardiac scans where a one-dimensional respiratory navigator is interleaved with a high-resolution three-dimensional (3D) segmented imaging sequence. The onedimensional real-time navigator scan usually has a different flip angle, sequence type, repetition time, echo time, and spatial resolution and orientation compared with the 3D segmented cardiac scan, which is acquired over several cardiac cycles (4). A limitation of the current approach is that the respiratory navigator acquisition typically is embedded in the imaging sequence and has a rigid implementation with severe restrictions on the available imaging parameters. This makes straightforward extension to two-dimensional (2D) and 3D navigators with different k-space sampling, image resolution, and orientation compared with the 3D segmented highresolution sequence challenging. This rigid software design also renders navigator optimization as well as comparison between different types of navigators (eg, recently developed 2D and 3D image-based navigators) difficult (5–8). In this study, we describe a recently developed framework for interleaving scans and demonstrate its usefulness for comparing different trajectories for image-based respiratory navigation in free-breathing coronary MRA (CMRA). The proposed method was evaluated in healthy subjects. METHODS The scan interleaving framework was implemented on a Philips Achieva 1.5T (Philips Healthcare, Best, The Netherlands) clinical scanner. The study was approved by the Institutional Review Board, and all subjects provided written informed consent. Scan Interleaving The proposed scan interleaving framework allowed for switching between independently defined scans at any
692
Interleaved Scanning for Cardiac MR
693
Table 1 Algorithm of Pseudocode for Scan Interleaving Navigator Scan
Segmented High-Resolution Scan
1. Start scan A 2. FOR each dynamic in scan A 3. scan_handover ( ) 4. Continue here after switch back 5. FOR each cardiac cycle in dynamic 6. FOR each profile in cycle 7. Readout scan A 8. END FOR 9. END FOR 10. END FOR 11. End scan A
time during the cardiovascular MR examination using a generic function call, scan_handover(), in the scanner software. Consider two scans in a cardiovascular MR protocol, scan A and scan B, with independent imaging parameters including location, orientation, and receive coils. During the execution of the pulse sequence of scan A, if scan_handover() is called, the progression of scan A will be stored and scan B will instantaneously start while scan A is suspended. Similarly, as soon as scan_handover() is called during the execution of scan B, its progression will be stored and scan A will continue at the point where it was previously suspended. In this fashion, the scans are interleaved until one scan finishes, at which point the remaining part of the unfinished scan will be performed without interleaving. In theory, any number of scans could be simultaneously interleaved using the scan interleaving framework. However, in practice, the amount of scans that could be concurrently interleaved was limited by the memory allocation capacity of the reconstruction computer. Scan Interleaving for Image-Based Respiratory Navigation From a software design perspective, a cardiovascular MR pulse sequence can be considered as a number of nested loops. The innermost loop performs the number of readouts specified by the k-space segment size. Other loops (from inner to outer) include looping over the cardiac phases and cardiac cycles, to name a few. With this software design, scan interleaving (as described in the previous section) can be achieved by inserting calls to scan_handover() into the appropriate loop to achieve
FIG. 1. Dynamic real-time 2D image navigators using a radial, spiral, or Cartesian EPI k-space trajectory interleaved with a segmented high-resolution 3D CMRA scan using Cartesian sampling. Prepulses (PP) included fat suppression and T2 preparation.
1. Start scan B 2. FOR each dynamic in scan B 3. FOR each cardiac cycle in dynamic 4. scan_handover ( ) 5. Continue here after switch back 6. FOR each profile in cycle 7. Readout scan B 8. END FOR 9. END FOR 10. END FOR 11. End scan B
interleaving at, for example, a dynamic scan level in the case of image-based navigation [where one dynamic scan refers to the acquisition of one image in a serial acquisition as often used in dynamic contrast-enhanced MRA (1)] or cardiac cycle in the case of a segmented k-space high-resolution scan. Table 1 shows the pseudocode for interleaving two scans, A (navigator scan) and B (segmented high-resolution scan), using nested loops. In this case, scan_handover() is called for each dynamic of scan A, whereas one cardiac cycle of scan B is performed every time before scan_handover() is called for scan B. With this framework, only the timing of the scan interleaving in the nested loops needs to be predefined during the software development stage. However, imaging parameters such as k-space trajectory, number of encoding dimensions, and spatial and temporal resolution for the interleaved scans remain arbitrary until the scan planning stage. MRI Experiments The proposed framework was used to interleave an electrocardiography-triggered segmented k-space high-resolution 3D whole heart CMRA scan with a 2D real-time dynamic scan, which was used for respiratory navigation. For the CMRA scan, one k-space segment (20–30 lines in k-space) was acquired per interleaf, while one 2D real-time image was acquired per navigator interleaf. For the navigator scan, radial (r-NAV), spiral (s-NAV), and Cartesian echo-planar imaging (CEPI-NAV) sampling trajectories were performed in three separate CMRA scans, while the 3D CMRA scan used a Cartesian trajectory (Fig. 1). The order of the three navigated scans was randomized for each subject. While the imaging parameters such as sampling density, repetition time, and echo time were different for the three navigator acquisitions, navigator temporal and spatial resolution, as well as all 3D CMRA imaging parameters were maintained in all three acquisitions. The imaging parameters for the 2D r-NAV, s-NAV, CEPI-NAV, and 3D whole heart CMRA scans are summarized in Table 2. A SENSE factor of two was used for the CMRA acquisition, and the nominal CMRA scan time was 4 min, 39 s, assuming a heart rate of 60 bpm. Because the study was designed to evaluate the ability of different navigator k-space trajectories to correct for respiratory motion of the heart, no respiratory gating was performed.
694
Henningsson et al.
Table 2 Imaging Parameters for r-NAV, s-NAV, CEPI-NAV, and CMRA Acquisitions Imaging Parameters Scan Type r-NAV s-NAV CEPI-NAV CMRA
Trajectory Radial Spiral Cartesian EPI Cartesian
Field of View (mm) 260 260 280 280 300 300 300 300 100
Spatial Resolution (mm) 2.5 2.5 2.5 2.5 2.5 2.5 1.3 1.3 1.3
Contrast GRE GRE GRE bSSFP
Flip Angle 10 10 10 70
Turbo Field Echo Factor 20 4 5 30
Sampling 20% 100% 100% 50%
Acquisition Time (ms) 61.9 68.6 62.8 124.8
bSSFP, balanced steady state free precession; GRE, spoiled gradient echo.
To allow for direct translation of the navigator motion estimation in foot–head and left–right direction to the CMRA motion correction, the image navigators and the CMRA acquisition had the same coronal orientation. The respiratory motion correction was performed retrospectively and involved translational correction of the CMRA data in foot–head and left–right directions by applying a linear phase to the acquired CMRA raw data based on the navigator motion estimation. Respiratory motion was extracted from the navigator images using a twodimensional normalized cross-correlation algorithm where the first navigator image was selected as the reference image and registered to all subsequent navigator images. Eight healthy subjects (mean age, 28 6 4; n ¼ 3 females, n ¼ 5 males) were scanned with the proposed CMRA protocol using r-NAV, s-NAV, and CEPI-NAV for respiratory motion correction. Data Analysis Navigator signal-to-noise ratio (SNR) is correlated with motion correction efficacy (9). An apparent SNR (aSNR) ratio, as described by Kolbitsch et al. (10), was calculated to compare the SNR of the different image navigators. To
this end, regions of interest were selected from the left ventricular blood pool and the air in one random navigator image for each navigated scan and healthy volunteer. To evaluate the motion correction performance of the navigators, vessel sharpness measurements were performed on the right coronary artery (RCA) and left anterior descending artery (LADA) of all acquired CMRA datasets using dedicated software (11). Additionally, a visual score from 0 (poor) to 4 (excellent) was given by an expert, blinded to the motion correction method used. The RCA and LADA were scored separately, and the order of the three motion correction approaches was randomized for each subject. Quantitative differences between navigator approaches were tested for statistical significance using a paired two-tailed t test where P < 0.05 was considered statistically significant. The qualitative visual score measurements were compared using a Wilcoxon signed-rank test, also with a significance threshold of P < 0.05. RESULTS Figure 2 shows representative respiratory navigator images throughout one respiratory cycle using r-NAV,
FIG. 2. Two-dimensional respiratory navigator images acquired using r-NAV, s-NAV, and CEPINAV sampling throughout one respiratory cycle of one healthy subject. The cross-hairs demonstrate the location of the calculated respiratory motion in the foot–head and left–right directions.
Interleaved Scanning for Cardiac MR
695
s-NAV, and CEPI-NAV. Crosshairs were added retrospectively to the navigator images in Fig. 2 to display the calculated respiratory position in the foot–head and left– right directions. The aSNR of the navigator images (mean 6 standard deviation) were as follows: r-NAV ¼ 12.3 6 1.9, s-NAV ¼ 24.5 6 5.0, and CEPI-NAV ¼ 28.1 6 11.1. The differences in aSNR between r-NAV and s-NAV were statistically significant (P < 0.05), as were the aSNR differences between r-NAV and CEPI-NAV (P < 0.01). The results from the visual score and vessel sharpness measurements for the RCA and LADA are summarized in Fig. 3. Higher visual scores and vessel sharpness were obtained using CEPI-NAV compared with r-NAV, comparisons for both the RCA and LADA yielding statistically significant differences (P < 0.05). A higher visual score was obtained for the LADA using s-NAV compared with r-NAV, which was statistically significant (P < 0.05). Furthermore, s-NAV yielded improved vessel sharpness compared with r-NAV, both for the RCA (P < 0.05) and LADA (P < 0.05). Although there was a general trend toward improved vessel sharpness and visual score using the CEPI-NAV compared with the s-NAV, no statistically significant differences were obtained. Figure 4 shows representative whole heart images from two healthy volunteers, reformatted to visualize the RCA and LADA, using either r-NAV, s-NAV, or C-NAV for respiratory motion correction, as well as an uncorrected scan. DISCUSSION FIG. 3. Quantitative and qualitative results of visual score and vessel sharpness measurements for the RCA and LADA. *P < 0.05.
In this study, we proposed a new framework for interleaving two or more scans, and investigated the feasibility of using it for image-based respiratory motion
FIG. 4. Whole heart CMRA from two healthy subjects reformatted to visualize the RCA and LADA, using no respiratory motion correction (uncorrected) and respiratory motion correction with r-NAV, s-NAV, or CEPI-NAV acquisition.
696
correction. From a pulse programming perspective, the proposed framework facilitates the implementation of scan interleaving. Previously, this was often achieved by hard-coding the pulse sequence of one of the interleaved scans into the code of another interleaved scan. The proposed approach allows for more flexible scan interleaving, as the interleaved scans can be independently defined, and the interleaving can be performed at an arbitrary level (eg, interleaving cardiac phases, cycles, scan dynamics, etc.) using a single generic function call. This allows for a more direct comparison of motion correction efficacy of different respiratory navigator trajectories, as all imaging parameters for the motion-corrected CMRA scan remain constant across all scans. The quantitative and qualitative comparisons between the whole heart CMRA acquired with r-NAV, s-NAV, or CEPI-NAV respiratory motion correction show an improved performance of s-NAV and CEPI-NAV compared with r-NAV. This is likely due to the higher SNR of spiral and Cartesian EPI trajectories compared with the radial trajectory. As we could show, the apparent SNR of the s-NAV and CEPI-NAV were significantly higher than the r-NAV by a factor of 2. A previous study has demonstrated the importance of navigator SNR on the respiratory motion correction efficacy using a one-dimensional pencil beam navigator (9). The findings of this study suggest a similar correlation, also for imagebased navigation. The high aSNR of the s-NAV and CEPINAV implies that such image navigators may be suitable for cardiovascular MR applications with an inherently lower SNR, such as late gadolinium enhancement or black-blood imaging. Further studies are required to evaluate the performance of image-based navigation in these applications. However, using the proposed framework for scan interleaving the implementation of such cardiovascular MR pulse sequences can be more readily achieved. A limitation of scan interleaving is the potential adverse influence of spin history effects between the different interleaved scans, such as magnetization saturation. However, this is a general problem for any interleaved scans, and not specific to the proposed framework. In this context, the proposed framework allows for simplified optimization of such protocols to minimize cross-talk artifacts. For the image-based navigators, reducing saturation effects on the CMRA acquisition could be achieved for the radial navigator by increasing the undersampling factor, for the spiral navigator by reducing the number of spiral interleaves (and increasing the repetition time), and for the Cartesian EPI navigator increasing the EPI factor. However, in our experiments, no visual saturation effects were observed in the CMRA acquisitions. To minimize saturation effects, particularly in the context of cardiovascular MR applications with
Henningsson et al.
inherently low SNR, navigator k-space trajectories such as EPI or spiral trajectories may be particularly good candidates because less magnetization is spent on the navigator acquisition. A limitation of this study is that only a 1.5T scanner system was used for data acquisition, and comparison of navigator trajectories. For higher field strengths such as 3T, which have less homogenous magnetic fields, EPI and spiral trajectories are more susceptible to image artifacts which may render them suboptimal compared with radial and standard Cartesian trajectories. Further studies are required to determine the optimal image navigator for CMRA at higher field strengths. In conclusion, a new scan interleaving framework has been developed, and its usefulness has been demonstrated for evaluating image-based respiratory motion correction with radial, spiral, and Cartesian EPI image navigators in coronary MRA. Using scan interleaving, we have shown that EPI and spiral navigator trajectories improve respiratory motion correction for coronary MRA compared with the radial navigator trajectory.
REFERENCES 1. Nael K, Laub G, Finn JP. Three-dimensional contrast-enhanced MR angiography of the thoraco-abdominal vessels. Magn Reson Imaging Clin N Am 2005;13:359–380. 2. Kellman P, Arai AE, McVeigh ER, Aletras AH. Phase-sensitive inversion recovery for detecting myocardial infarction using gadoliniumdelayed hyperenhancement. Magn Reson Med 2002;47:372–383. 3. Kozerke S, Tsao J, Razavi R, Boesiger P. Accelerating cardiac cine 3D imaging using k-t BLAST. Magn Reson Med 2004;52:19–26. 4. Danias PG, McConnell MV, Khasgiwala VC, Chuang ML, Edelman RR, Manning WJ. Prospective navigator correction of image position for coronary MR angiography. Radiology 1997;203:733–736. 5. Keegan J, Gatehouse PD, Yang GZ, Firmin DN. Non-model-based correction of respiratory motion using beat-to-beat 3D spiral fat-selective imaging. J Magn Reson Imaging 2007;26:624–629. 6. Henningsson M, Koken P, Stehning C, Razavi R, Prieto C, Botnar RM. Whole-heart coronary MR angiography with 2D self-navigated image reconstruction. Magn Reson Med 2012;67:437–445. 7. Wu HH, Gurney PT, Hu BS, Nishimura DG, McConnell MV. Freebreathing multiphase whole-heart coronary MR angiography using image-based navigators and three-dimensional cones imaging. Magn Reson Med 2013;69:1083–1093. 8. Kawaji K, Spincemaille P, Nguyen TD, Thimmappa N, Cooper MA, Prince MR, Wang Y. Direct coronary motion extraction from a 2D fat image navigator for prospectively gated coronary MR angiography. Magn Reson Med 2014;71:599–607. 9. Spuentrup E, Stuber M, Botnar RM, Manning WJ. The impact of navigator timing parameters and navigator spatial resolution on 3D coronary magnetic resonance angiography. J Magn Reson Imaging 2001; 14:311–318. 10. Kolbitsch C, Prieto C, Smink J, Schaeffter T. Highly efficient wholeheart imaging using radial phase encoding-phase ordering with automatic window selection. Magn Reson Med 2011;66:1008–1018. 11. Etienne A, Botnar RM, Van Muiswinkel AM, Boesiger P, Manning WJ, Stuber M. “Soap-Bubble” visualization and quantitative analysis of 3D coronary magnetic resonance angiograms. Magn Reson Med 2002;48:658–666.
A New Framework for Interleaved Scanning in Cardiovascular MR: Application to Image-Based Respiratory Motion Correction in Coronary MR Angiography Markus Henningsson,1,2,3,4* Giel Mens,5 Peter Koken,6 Jouke Smink,5 and Rene M. Botnar1,2,3,4 INTRODUCTION
Purpose: To describe a new framework for interleaving scans and demonstrate its usefulness for image-based respiratory motion correction in whole heart coronary MR angiography (CMRA). Methods: Scan interleaving using the proposed approach was achieved by switching between separately defined, independent scans at arbitrary time points during their execution, using a generic function call. The scan interleaving framework was used to perform scan interleaving for image-based respiratory navigation of CMRA with spiral, radial, and Cartesian echoplanar imaging (EPI) navigator k-space trajectories. Eight healthy volunteers were scanned. Results: Improved coronary vessel sharpness and visual scores were obtained using spiral and Cartesian EPI navigators compared with radial navigators. Conclusion: The usefulness of the proposed scan interleaving framework was demonstrated for image-based respiratory motion correction. It facilitated more direct comparisons of image navigator acquisitions with different k-space trajectories. Furthermore, we could demonstrate that spiral and Cartesian EPI navigators may be particularly suitable for image-based motion correction, as they provide improved motion correction and high navigator apparent signal-to-noise ratio while spending very little magnetization, thereby minimizing saturation C 2014 Wiley effects. Magn Reson Med 73:692–696, 2015. V Periodicals, Inc. Key words: interleaved scanning; cardiovascular MRI; respiratory motion correction
1 Division of Imaging Sciences & Biomedical Engineering, King’s College London, London, UK. 2 Wellcome Trust and EPSRC Medical Engineering Center, King’s College London, London, UK. 3 BHF Centre of Excellence, King’s College London, London, UK. 4 NIHR Biomedical Research Centre, King’s College London, London, UK.. 5 Philips Healthcare, Best, The Netherlands. 6 Philips Research, Hamburg, Germany
*Correspondence to: Markus Henningsson, Ph.D., King’s College London, Division of Imaging Sciences and Biomedical Engineering, The Rayne Institute, 4th Floor, Lambeth Wing, St. Thomas’ Hospital, London, SE1 7EH, UK. E-mail: [email protected] Correction added after online publication 24 March 2014. Author corrected the coauthor affiliation numbering and the e-mail address of the corresponding author. Received 11 June 2013; revised 6 January 2014; accepted 7 January 2014 DOI 10.1002/mrm.25149 Published online 17 March 2014 in Wiley Online Library (wileyonlinelibrary. com). C 2014 Wiley Periodicals, Inc. V
For a long time there has been a need to interleave concurrently performed scans with more or less independent imaging parameters in cardiovascular MR, including bolus tracking in contrast-enhanced MR angiography (MRA) (1), phase-sensitive inversion recovery late gadolinium enhancement (2), or acceleration techniques for time-resolved scans such as k-t BLAST (3) that uses interleaved training data sets. Perhaps the best example of interleaved scanning in cardiovascular MR is in free-breathing navigator-gated cardiac scans where a one-dimensional respiratory navigator is interleaved with a high-resolution three-dimensional (3D) segmented imaging sequence. The onedimensional real-time navigator scan usually has a different flip angle, sequence type, repetition time, echo time, and spatial resolution and orientation compared with the 3D segmented cardiac scan, which is acquired over several cardiac cycles (4). A limitation of the current approach is that the respiratory navigator acquisition typically is embedded in the imaging sequence and has a rigid implementation with severe restrictions on the available imaging parameters. This makes straightforward extension to two-dimensional (2D) and 3D navigators with different k-space sampling, image resolution, and orientation compared with the 3D segmented highresolution sequence challenging. This rigid software design also renders navigator optimization as well as comparison between different types of navigators (eg, recently developed 2D and 3D image-based navigators) difficult (5–8). In this study, we describe a recently developed framework for interleaving scans and demonstrate its usefulness for comparing different trajectories for image-based respiratory navigation in free-breathing coronary MRA (CMRA). The proposed method was evaluated in healthy subjects. METHODS The scan interleaving framework was implemented on a Philips Achieva 1.5T (Philips Healthcare, Best, The Netherlands) clinical scanner. The study was approved by the Institutional Review Board, and all subjects provided written informed consent. Scan Interleaving The proposed scan interleaving framework allowed for switching between independently defined scans at any
692
Interleaved Scanning for Cardiac MR
693
Table 1 Algorithm of Pseudocode for Scan Interleaving Navigator Scan
Segmented High-Resolution Scan
1. Start scan A 2. FOR each dynamic in scan A 3. scan_handover ( ) 4. Continue here after switch back 5. FOR each cardiac cycle in dynamic 6. FOR each profile in cycle 7. Readout scan A 8. END FOR 9. END FOR 10. END FOR 11. End scan A
time during the cardiovascular MR examination using a generic function call, scan_handover(), in the scanner software. Consider two scans in a cardiovascular MR protocol, scan A and scan B, with independent imaging parameters including location, orientation, and receive coils. During the execution of the pulse sequence of scan A, if scan_handover() is called, the progression of scan A will be stored and scan B will instantaneously start while scan A is suspended. Similarly, as soon as scan_handover() is called during the execution of scan B, its progression will be stored and scan A will continue at the point where it was previously suspended. In this fashion, the scans are interleaved until one scan finishes, at which point the remaining part of the unfinished scan will be performed without interleaving. In theory, any number of scans could be simultaneously interleaved using the scan interleaving framework. However, in practice, the amount of scans that could be concurrently interleaved was limited by the memory allocation capacity of the reconstruction computer. Scan Interleaving for Image-Based Respiratory Navigation From a software design perspective, a cardiovascular MR pulse sequence can be considered as a number of nested loops. The innermost loop performs the number of readouts specified by the k-space segment size. Other loops (from inner to outer) include looping over the cardiac phases and cardiac cycles, to name a few. With this software design, scan interleaving (as described in the previous section) can be achieved by inserting calls to scan_handover() into the appropriate loop to achieve
FIG. 1. Dynamic real-time 2D image navigators using a radial, spiral, or Cartesian EPI k-space trajectory interleaved with a segmented high-resolution 3D CMRA scan using Cartesian sampling. Prepulses (PP) included fat suppression and T2 preparation.
1. Start scan B 2. FOR each dynamic in scan B 3. FOR each cardiac cycle in dynamic 4. scan_handover ( ) 5. Continue here after switch back 6. FOR each profile in cycle 7. Readout scan B 8. END FOR 9. END FOR 10. END FOR 11. End scan B
interleaving at, for example, a dynamic scan level in the case of image-based navigation [where one dynamic scan refers to the acquisition of one image in a serial acquisition as often used in dynamic contrast-enhanced MRA (1)] or cardiac cycle in the case of a segmented k-space high-resolution scan. Table 1 shows the pseudocode for interleaving two scans, A (navigator scan) and B (segmented high-resolution scan), using nested loops. In this case, scan_handover() is called for each dynamic of scan A, whereas one cardiac cycle of scan B is performed every time before scan_handover() is called for scan B. With this framework, only the timing of the scan interleaving in the nested loops needs to be predefined during the software development stage. However, imaging parameters such as k-space trajectory, number of encoding dimensions, and spatial and temporal resolution for the interleaved scans remain arbitrary until the scan planning stage. MRI Experiments The proposed framework was used to interleave an electrocardiography-triggered segmented k-space high-resolution 3D whole heart CMRA scan with a 2D real-time dynamic scan, which was used for respiratory navigation. For the CMRA scan, one k-space segment (20–30 lines in k-space) was acquired per interleaf, while one 2D real-time image was acquired per navigator interleaf. For the navigator scan, radial (r-NAV), spiral (s-NAV), and Cartesian echo-planar imaging (CEPI-NAV) sampling trajectories were performed in three separate CMRA scans, while the 3D CMRA scan used a Cartesian trajectory (Fig. 1). The order of the three navigated scans was randomized for each subject. While the imaging parameters such as sampling density, repetition time, and echo time were different for the three navigator acquisitions, navigator temporal and spatial resolution, as well as all 3D CMRA imaging parameters were maintained in all three acquisitions. The imaging parameters for the 2D r-NAV, s-NAV, CEPI-NAV, and 3D whole heart CMRA scans are summarized in Table 2. A SENSE factor of two was used for the CMRA acquisition, and the nominal CMRA scan time was 4 min, 39 s, assuming a heart rate of 60 bpm. Because the study was designed to evaluate the ability of different navigator k-space trajectories to correct for respiratory motion of the heart, no respiratory gating was performed.
694
Henningsson et al.
Table 2 Imaging Parameters for r-NAV, s-NAV, CEPI-NAV, and CMRA Acquisitions Imaging Parameters Scan Type r-NAV s-NAV CEPI-NAV CMRA
Trajectory Radial Spiral Cartesian EPI Cartesian
Field of View (mm) 260 260 280 280 300 300 300 300 100
Spatial Resolution (mm) 2.5 2.5 2.5 2.5 2.5 2.5 1.3 1.3 1.3
Contrast GRE GRE GRE bSSFP
Flip Angle 10 10 10 70
Turbo Field Echo Factor 20 4 5 30
Sampling 20% 100% 100% 50%
Acquisition Time (ms) 61.9 68.6 62.8 124.8
bSSFP, balanced steady state free precession; GRE, spoiled gradient echo.
To allow for direct translation of the navigator motion estimation in foot–head and left–right direction to the CMRA motion correction, the image navigators and the CMRA acquisition had the same coronal orientation. The respiratory motion correction was performed retrospectively and involved translational correction of the CMRA data in foot–head and left–right directions by applying a linear phase to the acquired CMRA raw data based on the navigator motion estimation. Respiratory motion was extracted from the navigator images using a twodimensional normalized cross-correlation algorithm where the first navigator image was selected as the reference image and registered to all subsequent navigator images. Eight healthy subjects (mean age, 28 6 4; n ¼ 3 females, n ¼ 5 males) were scanned with the proposed CMRA protocol using r-NAV, s-NAV, and CEPI-NAV for respiratory motion correction. Data Analysis Navigator signal-to-noise ratio (SNR) is correlated with motion correction efficacy (9). An apparent SNR (aSNR) ratio, as described by Kolbitsch et al. (10), was calculated to compare the SNR of the different image navigators. To
this end, regions of interest were selected from the left ventricular blood pool and the air in one random navigator image for each navigated scan and healthy volunteer. To evaluate the motion correction performance of the navigators, vessel sharpness measurements were performed on the right coronary artery (RCA) and left anterior descending artery (LADA) of all acquired CMRA datasets using dedicated software (11). Additionally, a visual score from 0 (poor) to 4 (excellent) was given by an expert, blinded to the motion correction method used. The RCA and LADA were scored separately, and the order of the three motion correction approaches was randomized for each subject. Quantitative differences between navigator approaches were tested for statistical significance using a paired two-tailed t test where P < 0.05 was considered statistically significant. The qualitative visual score measurements were compared using a Wilcoxon signed-rank test, also with a significance threshold of P < 0.05. RESULTS Figure 2 shows representative respiratory navigator images throughout one respiratory cycle using r-NAV,
FIG. 2. Two-dimensional respiratory navigator images acquired using r-NAV, s-NAV, and CEPINAV sampling throughout one respiratory cycle of one healthy subject. The cross-hairs demonstrate the location of the calculated respiratory motion in the foot–head and left–right directions.
Interleaved Scanning for Cardiac MR
695
s-NAV, and CEPI-NAV. Crosshairs were added retrospectively to the navigator images in Fig. 2 to display the calculated respiratory position in the foot–head and left– right directions. The aSNR of the navigator images (mean 6 standard deviation) were as follows: r-NAV ¼ 12.3 6 1.9, s-NAV ¼ 24.5 6 5.0, and CEPI-NAV ¼ 28.1 6 11.1. The differences in aSNR between r-NAV and s-NAV were statistically significant (P < 0.05), as were the aSNR differences between r-NAV and CEPI-NAV (P < 0.01). The results from the visual score and vessel sharpness measurements for the RCA and LADA are summarized in Fig. 3. Higher visual scores and vessel sharpness were obtained using CEPI-NAV compared with r-NAV, comparisons for both the RCA and LADA yielding statistically significant differences (P < 0.05). A higher visual score was obtained for the LADA using s-NAV compared with r-NAV, which was statistically significant (P < 0.05). Furthermore, s-NAV yielded improved vessel sharpness compared with r-NAV, both for the RCA (P < 0.05) and LADA (P < 0.05). Although there was a general trend toward improved vessel sharpness and visual score using the CEPI-NAV compared with the s-NAV, no statistically significant differences were obtained. Figure 4 shows representative whole heart images from two healthy volunteers, reformatted to visualize the RCA and LADA, using either r-NAV, s-NAV, or C-NAV for respiratory motion correction, as well as an uncorrected scan. DISCUSSION FIG. 3. Quantitative and qualitative results of visual score and vessel sharpness measurements for the RCA and LADA. *P < 0.05.
In this study, we proposed a new framework for interleaving two or more scans, and investigated the feasibility of using it for image-based respiratory motion
FIG. 4. Whole heart CMRA from two healthy subjects reformatted to visualize the RCA and LADA, using no respiratory motion correction (uncorrected) and respiratory motion correction with r-NAV, s-NAV, or CEPI-NAV acquisition.
696
correction. From a pulse programming perspective, the proposed framework facilitates the implementation of scan interleaving. Previously, this was often achieved by hard-coding the pulse sequence of one of the interleaved scans into the code of another interleaved scan. The proposed approach allows for more flexible scan interleaving, as the interleaved scans can be independently defined, and the interleaving can be performed at an arbitrary level (eg, interleaving cardiac phases, cycles, scan dynamics, etc.) using a single generic function call. This allows for a more direct comparison of motion correction efficacy of different respiratory navigator trajectories, as all imaging parameters for the motion-corrected CMRA scan remain constant across all scans. The quantitative and qualitative comparisons between the whole heart CMRA acquired with r-NAV, s-NAV, or CEPI-NAV respiratory motion correction show an improved performance of s-NAV and CEPI-NAV compared with r-NAV. This is likely due to the higher SNR of spiral and Cartesian EPI trajectories compared with the radial trajectory. As we could show, the apparent SNR of the s-NAV and CEPI-NAV were significantly higher than the r-NAV by a factor of 2. A previous study has demonstrated the importance of navigator SNR on the respiratory motion correction efficacy using a one-dimensional pencil beam navigator (9). The findings of this study suggest a similar correlation, also for imagebased navigation. The high aSNR of the s-NAV and CEPINAV implies that such image navigators may be suitable for cardiovascular MR applications with an inherently lower SNR, such as late gadolinium enhancement or black-blood imaging. Further studies are required to evaluate the performance of image-based navigation in these applications. However, using the proposed framework for scan interleaving the implementation of such cardiovascular MR pulse sequences can be more readily achieved. A limitation of scan interleaving is the potential adverse influence of spin history effects between the different interleaved scans, such as magnetization saturation. However, this is a general problem for any interleaved scans, and not specific to the proposed framework. In this context, the proposed framework allows for simplified optimization of such protocols to minimize cross-talk artifacts. For the image-based navigators, reducing saturation effects on the CMRA acquisition could be achieved for the radial navigator by increasing the undersampling factor, for the spiral navigator by reducing the number of spiral interleaves (and increasing the repetition time), and for the Cartesian EPI navigator increasing the EPI factor. However, in our experiments, no visual saturation effects were observed in the CMRA acquisitions. To minimize saturation effects, particularly in the context of cardiovascular MR applications with
Henningsson et al.
inherently low SNR, navigator k-space trajectories such as EPI or spiral trajectories may be particularly good candidates because less magnetization is spent on the navigator acquisition. A limitation of this study is that only a 1.5T scanner system was used for data acquisition, and comparison of navigator trajectories. For higher field strengths such as 3T, which have less homogenous magnetic fields, EPI and spiral trajectories are more susceptible to image artifacts which may render them suboptimal compared with radial and standard Cartesian trajectories. Further studies are required to determine the optimal image navigator for CMRA at higher field strengths. In conclusion, a new scan interleaving framework has been developed, and its usefulness has been demonstrated for evaluating image-based respiratory motion correction with radial, spiral, and Cartesian EPI image navigators in coronary MRA. Using scan interleaving, we have shown that EPI and spiral navigator trajectories improve respiratory motion correction for coronary MRA compared with the radial navigator trajectory.
REFERENCES 1. Nael K, Laub G, Finn JP. Three-dimensional contrast-enhanced MR angiography of the thoraco-abdominal vessels. Magn Reson Imaging Clin N Am 2005;13:359–380. 2. Kellman P, Arai AE, McVeigh ER, Aletras AH. Phase-sensitive inversion recovery for detecting myocardial infarction using gadoliniumdelayed hyperenhancement. Magn Reson Med 2002;47:372–383. 3. Kozerke S, Tsao J, Razavi R, Boesiger P. Accelerating cardiac cine 3D imaging using k-t BLAST. Magn Reson Med 2004;52:19–26. 4. Danias PG, McConnell MV, Khasgiwala VC, Chuang ML, Edelman RR, Manning WJ. Prospective navigator correction of image position for coronary MR angiography. Radiology 1997;203:733–736. 5. Keegan J, Gatehouse PD, Yang GZ, Firmin DN. Non-model-based correction of respiratory motion using beat-to-beat 3D spiral fat-selective imaging. J Magn Reson Imaging 2007;26:624–629. 6. Henningsson M, Koken P, Stehning C, Razavi R, Prieto C, Botnar RM. Whole-heart coronary MR angiography with 2D self-navigated image reconstruction. Magn Reson Med 2012;67:437–445. 7. Wu HH, Gurney PT, Hu BS, Nishimura DG, McConnell MV. Freebreathing multiphase whole-heart coronary MR angiography using image-based navigators and three-dimensional cones imaging. Magn Reson Med 2013;69:1083–1093. 8. Kawaji K, Spincemaille P, Nguyen TD, Thimmappa N, Cooper MA, Prince MR, Wang Y. Direct coronary motion extraction from a 2D fat image navigator for prospectively gated coronary MR angiography. Magn Reson Med 2014;71:599–607. 9. Spuentrup E, Stuber M, Botnar RM, Manning WJ. The impact of navigator timing parameters and navigator spatial resolution on 3D coronary magnetic resonance angiography. J Magn Reson Imaging 2001; 14:311–318. 10. Kolbitsch C, Prieto C, Smink J, Schaeffter T. Highly efficient wholeheart imaging using radial phase encoding-phase ordering with automatic window selection. Magn Reson Med 2011;66:1008–1018. 11. Etienne A, Botnar RM, Van Muiswinkel AM, Boesiger P, Manning WJ, Stuber M. “Soap-Bubble” visualization and quantitative analysis of 3D coronary magnetic resonance angiograms. Magn Reson Med 2002;48:658–666.
