Challenges & Limitations in Ischemia Imaging
Ed DiBella1

1Department of Radiology and Imaging Sciences, University of Utah

Synopsis

Myocardial perfusion acquisitions have high sensitivity/specificity for the detection of ischemia though are challenged by motion, dark rim artifact, and issues with quantification. These issues are briefly addressed in this syllabus, with references to more of the work done in these areas.

Introduction

Title of Session: Cardiac MRI: The Basic Principles & Applications

Speaker Name: Ed DiBella, edward.dibella@hsc.utah.edu

Talk Title: Challenges and Limitations in Ischemia Imaging

Target audience: Researchers and clinicians interested in ischemia imaging using cardiac MRI.

Objectives: – After this presentation, one will be able to 1) list three challenges to performing standard myocardial perfusion imaging, and 2) discuss methods for handling motion and other artifacts.

PURPOSE: – To review challenges and limitations for myocardial perfusion MRI, and to present possible new solutions to some of the challenges with perfusion imaging.

Issues and work done in the field

Myocardial perfusion has become a very useful method for detecting and characterizing ischemia, with sensitivity/specificities of 90/80% in meta-analyses [1, 2] or 90/87% compared to FFR [3]. There is still room for improvement, as there are challenges such as dark rim artifact, respiratory motion, and cardiac motion. In addition, quantification has been improving, and can be particularly useful to assess balanced (triple-vessel) disease that can be difficult to detect otherwise.

Dark Rim Artifact The dark rim artifact can masquerade as subendocardial ischemia. Dark rim artifact can often be “read through” since it typically does not last for many time frames if it is caused by the “Gibbs ringing” from the bright blood pool edge [4]. Intra-frame motion can also play a role in dark rim artifacts [5]. Fig. 1 shows an example of the Gibbs ringing artifact and the effects of spatial resolution, in an ex-vivo preparation used to eliminate motion but mimic the arrival of the gadolinium bolus in the left ventricle blood pool. Higher spatial and temporal resolution, radial imaging, and spatial filtering (apodization) [6, 7] tend to result in much less or no dark rim artifacts.

Respiratory motion While breath-hold scans are ideal in that the only changes over time are due to gadolinium concentration changes, large breaths at inopportune times can be a serious drawback. Typically, not all subjects can hold their breath long enough for the first pass study. This is especially true during pharmacological stress, or (even more challenging) at exercise stress [8, 9]. Thus respiratory motion needs to be considered.

Free-breathing perfusion scans have been shown to have good sensitivity and specificity – this is likely because physicians reading the scans are good at ignoring inter-frame motion. Intra-frame motion is generally not an issue from respiration since most acquisitions are short (<150 msec) relative to respiratory changes. Out-of-plane motion can be a consideration but is typically relatively small in free-breathing scans. For quantification or for possibly improved reading, there are essentially three ways to handle inter-frame motion.

The first method for handling the inter-frame motion, and by far the most widely used, is to register the time frames post-acquisition. Both rigid and non-rigid motion compensation methods have been explored by numerous groups – see survey in [10], and papers such as [11-15]. The change in contrast over time complicates the registration compared to typical image registration methods, but has been addressed in several different ways. For example, mutual information [16], statistically based methods [17], iterative comparisons with a reference image [18], model-based methods [19], and independent component analysis (ICA) [20] have all been formulated to work for the time-varying contrast in the images. Some of these groups have provided data in order for others to compare alternative methods on the same datasets. For example, Wollny et al. compared methods and provided data online for others to compare in [21]. And a 2014 MICCAI STACOM workshop ran a perfusion registration challenge in which different groups compared their methods directly on common datasets (see http://www.cardiacatlas.org/web/stacom2014/moco-introduction and results at https://avan.shinyapps.io/MoCoSTACOM2014/). [22] The large number of papers for image registration of perfusion images reflects that this is not a solved problem for some datasets, although images with high SNR and relatively small motion (shallow breathing) seem to be readily registered with a number of the published techniques.

A second approach for handling inter-frame motion is relatively new. This is to include motion compensation within the image reconstruction. In particular, for high undersampling rates and for reconstructions leveraging temporal correlations, estimating motion should provide significant gains in image quality. These gains can exist whether or not motion free images are reconstructed. Examples of such approaches applied to perfusion include [23-30]. These methods are generally very computationally intensive, though it makes intuitive sense not to decouple the image reconstruction and registration processes as heavily undersampled acquisitions become the norm.

A third approach is to handle inter-frame motion by modifying the acquisition to include prospective slice-tracking. Only a few groups have reported on this approach [9, 31, 32]. An extra navigator readout is performed for each time frame and used to adjust the position of the slice in that time frame. This approach can be particularly helpful for exercise stress studies, where breathing is deep and rapid, causing significant out-of-plane motion [9]. Obtaining a robust navigator signal can be challenging at 3T [31].

What about respiratory motion issues for 3D perfusion acquisitions? More 3D perfusion studies are being published, mostly using breath-hold acquisitions. Similar issues as with 2D arise, although the fact that each 3D set of slices has the same contrast and the same timing (cardiac phase and respiratory phase) at each heartbeat makes the problem of registration easier in principle. Through- plane motion too is different in that it is likely correctable in central slices since it essentially changes only edges of the slab, and only one slab edge would have new data.

Cardiac Motion With good ECG-gating, and fast readouts, cardiac motion is frozen and not an issue. ECG-gating is still problematic for some studies, especially at 3T. Additionally, arrhythmias are a concern that typically result in missed beats and inefficient acquisitions. These problems do not only affect perfusion scans. For example, cine scans can have gating/rhythm issues, and thus a number of groups have developed real-time or self-gated cine studies. Similar approaches are possible for perfusion, since with current advances, very rapid readouts are possible - a block of slices can be acquired repeatedly, without any gating signals. These ungated images can then be read by physicians [33], or the frames can be retrospectively binned (“self-gated”) into systole or diastole prior to reading [33] or for quantitative analysis [34, 35]. With this type of ungated free-breathing acquisition, perfusion becomes a much simpler protocol to prescribe and run on the scanner.

Ungated approaches also naturally lead to the idea of dropping the saturation pulse and acquiring 3D acquisitions at steady state [36, 37], much like 3D tumor DCE scans are done with spoiled gradient echo (SPGR) sequences. Other alternatives such as 3D SSFP with T2/T1 contrast have been published [38]. And 2D SPGR methods with single-slice [39] and multi-slice acquisitions are beginning to be studied as well. There are trade-offs with these imaging approaches; it is not clear what approach is best but it is clear that perfusion imaging has exciting times ahead.

Quantification Quantifying myocardial perfusion in ml/min/g has the challenge of obtaining an accurate arterial input function since the signal from gadolinium saturates, becoming nonlinear at higher gadolinium doses. Methods such as using very small doses, a dual bolus [40], blind estimation [41], or dual sequence [42] can work to avoid the saturation issue. Tissue can be saturated as well with high doses or long saturation recovery times. In addition, the motion issues mentioned above can make quantification more complex and/or less reliable. Finally, the best model to use to obtain perfusion values is not well-accepted. Some investigators use the Fermi model, and others use a compartment model or Patlak model and include an extraction fraction term. More work is needed to determine the best approach and how much value it adds for assessing myocardial ischemia.

Conclusion

Myocardial perfusion acquisitions have high sensitivity/specificity though continue to improve towards achieving full myocardial coverage and quantification (to address for example triple vessel disease), and to be artifact-free. Free-breathing scans and even ungated acquisitions are being performed, and volume coverage is increasing via 3D and simultaneous multi-slice methods. In the future it is likely that acquisitions will become simpler and more efficient by moving more of the complexity to the image reconstruction and registration software. Such advances will widen use of the technique, increasing clinical impact.

Acknowledgements

We gratefully acknowledge support from the NIH towards addressing some of the challenges associated with MRI myocardial perfusion methods.

References

[1] K. R. Nandalur, B. A. Dwamena, A. F. Choudhri, M. R. Nandalur, and R. C. Carlos, "Diagnostic performance of stress cardiac magnetic resonance imaging in the detection of coronary artery disease: a meta-analysis," J Am Coll Cardiol, vol. 50, pp. 1343-53, 2007.

[2] M. Hamon, G. Fau, G. Nee, J. Ehtisham, and R. Morello, "Meta-analysis of the diagnostic performance of stress perfusion cardiovascular magnetic resonance for detection of coronary artery disease," J Cardiovasc Magn Reson, vol. 12, p. 29, 2010.

[3] M. Li, T. Zhou, L. F. Yang, Z. H. Peng, J. Ding, et al., "Diagnostic accuracy of myocardial magnetic resonance perfusion to diagnose ischemic stenosis with fractional flow reserve as reference: systematic review and meta-analysis," JACC Cardiovasc Imaging, vol. 7, pp. 1098-105, 2014.

[4] E. V. Di Bella, D. L. Parker, and A. J. Sinusas, "On the dark rim artifact in dynamic contrast-enhanced MRI myocardial perfusion studies," Magn Reson Med, vol. 54, pp. 1295-9, 2005.

[5] P. Storey, Q. Chen, W. Li, R. R. Edelman, and P. V. Prasad, "Band artifacts due to bulk motion," Magn Reson Med, vol. 48, pp. 1028-36, 2002.

[6] M. Salerno, C. Sica, C. M. Kramer, and C. H. Meyer, "Improved first-pass spiral myocardial perfusion imaging with variable density trajectories," Magn Reson Med, vol. 70, pp. 1369-79, 2013.

[7] B. Sharif, R. Dharmakumar, T. LaBounty, R. Arsanjani, C. Shufelt, et al., "Towards elimination of the dark-rim artifact in first-pass myocardial perfusion MRI: removing Gibbs ringing effects using optimized radial imaging," Magn Reson Med, vol. 72, pp. 124-36, 2014.

[8] E. L. Foster, J. W. Arnold, M. Jekic, J. A. Bender, V. Balasubramanian, et al., "MR-compatible treadmill for exercise stress cardiac magnetic resonance imaging," Magn Reson Med, vol. 67, pp. 880-9, 2012.

[9] T. A. Basha, S. Roujol, K. V. Kissinger, B. Goddu, S. Berg, et al., "Free-breathing cardiac MR stress perfusion with real-time slice tracking," Magn Reson Med, vol. 72, pp. 689-98, 2014.

[10] V. Gupta, H. A. Kirisli, E. A. Hendriks, R. J. van der Geest, M. van de Giessen, et al., "Cardiac MR perfusion image processing techniques: a survey," Med Image Anal, vol. 16, pp. 767-85, 2012.

[11] M. B. Stegmann, H. Olafsdottir, and H. B. Larsson, "Unsupervised motion-compensation of multi-slice cardiac perfusion MRI," Med Image Anal, vol. 9, pp. 394-410, 2005.

[12] H. Xue, S. Zuehlsdorff, P. Kellman, A. Arai, S. Nielles-Vallespin, et al., "Unsupervised inline analysis of cardiac perfusion MRI," Med Image Comput Comput Assist Interv, vol. 12, pp. 741-9, 2009.

[13] G. Tarroni, C. Corsi, P. F. Antkowiak, F. Veronesi, C. M. Kramer, et al., "Myocardial perfusion: near-automated evaluation from contrast-enhanced MR images obtained at rest and during vasodilator stress," Radiology, vol. 265, pp. 576-83, 2012.

[14] G. Wollny, M. J. Ledesma-Carbayo, P. Kellman, and A. Santos, "Exploiting quasiperiodicity in motion correction of free-breathing myocardial perfusion MRI," IEEE Trans Med Imaging, vol. 29, pp. 1516-27, 2010.

[15] G. Mihai, Y. Ding, H. Xue, Y. C. Chung, S. Rajagopalan, et al., "Non-rigid registration and KLT filter to improve SNR and CNR in GRE-EPI myocardial perfusion imaging," J Biomed Sci Eng, vol. 5, pp. 871-877, 2012.

[16] G. M. Beache, F. Khalifa, A. El-Baz, and G. Gimel'farb, "Fully automated framework for the analysis of myocardial first-pass perfusion MR images," Med Phys, vol. 41, p. 102305, 2014.

[17] C. M. Gallippi, C. M. Kramer, Y. L. Hu, D. A. Vido, N. Reichek, et al., "Fully automated registration and warping of contrast-enhanced first-pass perfusion images," J Cardiovasc Magn Reson, vol. 4, pp. 459-69, 2002.

[18] L. M. Bidaut and J. P. Vallee, "Automated registration of dynamic MR images for the quantification of myocardial perfusion," J Magn Reson Imaging, vol. 13, pp. 648-55, 2001. [19] G. Adluru, E. V. R. DiBella, and M. C. Schabel, "Model-based registration for dynamic cardiac perfusion MRI," J Magn Reson Imaging, vol. 24, pp. 1062-1070, 2006.

[20] G. Wollny, P. Kellman, A. Santos, and M. J. Ledesma-Carbayo, "Automatic motion compensation of free breathing acquired myocardial perfusion data by using independent component analysis," Med Image Anal, vol. 16, pp. 1015-28, 2012.

[21] G. Wollny and P. Kellman, "Free breathing myocardial perfusion data sets for performance analysis of motion compensation algorithms," Gigascience, vol. 3, p. 23, 2014.

[22] "Statistical Atlases and Computational Models of the Heart - Imaging and Modelling Challenges," in 5th International Workshop, STACOM 2014, Held in Conjunction with MICCAI Boston, MA, 2014.

[23] S. G. Lingala, M. Nadar, C. Chefd-Hotel, L. Zhang, and M. Jacob, "Unified reconstruction and motion estimation in cardiac perfusion MRI," in Int Soc Biomedical Imaging (ISBI), Chicago, 2011, pp. 65-68.

[24] G. Adluru and E. DiBella, "Reconstruction with diffeomorphic motion compensation for undersampled dynamic MRI," in Proc of SPIE, San Diego, CA, 2013, pp. 88581X, p.1-4.

[25] S. G. Lingala, E. DiBella, and M. Jacob, "Deformation Corrected Compressed Sensing (DC-CS): A Novel Framework for Accelerated Dynamic MRI," IEEE Trans Med Imaging, vol. 34, pp. 72-85, 2015.

[26] M. Filipovic, P. A. Vuissoz, A. Codreanu, M. Claudon, and J. Felblinger, "Motion compensated generalized reconstruction for free-breathing dynamic contrast-enhanced MRI," Magn Reson Med, vol. 65, pp. 812-22, 2011.

[27] L. Ge, A. Kino, M. Griswold, J. C. Carr, and D. Li, "Free-breathing myocardial perfusion MRI using SW-CG-HYPR and motion correction," Magn Reson Med, vol. 64, pp. 1148-54, 2010.

[28] R. Otazo, E. Candes, and D. K. Sodickson, "Low-rank plus sparse matrix decomposition for accelerated dynamic MRI with separation of background and dynamic components," Magn Reson Med, 2014.

[29] N. K. Naresh, X. Chen, R. J. Roy, P. F. Antkowiak, B. H. Annex, et al., "Accelerated dual-contrast first-pass perfusion MRI of the mouse heart: Development and application to diet-induced obese mice," Magn Reson Med, 2014.

[30] X. Chen, M. Salerno, Y. Yang, and F. H. Epstein, "Motion-compensated compressed sensing for dynamic contrast-enhanced MRI using regional spatiotemporal sparsity and region tracking: block low-rank sparsity with motion-guidance (BLOSM)," Magn Reson Med, vol. 72, pp. 1028-38, 2014.

[31] H. Pedersen, S. Kelle, S. Ringgaard, B. Schnackenburg, E. Nagel, et al., "Quantification of myocardial perfusion using free-breathing MRI and prospective slice tracking," Magn Reson Med, vol. 61, pp. 734-8, 2009.

[32] D. E. J. Cleppien, G. Horstick, N. Abegunewardene, S. Weber, C. E. Müller, et al., "Comparison of the quantitative first pass myocardial perfusion MRI with and without prospective slice tracking: Comparison between breath-hold and free-breathing condition," Magn Reson Med, vol. 64, pp. 1461-70, 2010.

[33] A. Harrison, G. Adluru, K. Damal, A. M. Shaaban, B. Wilson, et al., "Rapid ungated myocardial perfusion cardiovascular magnetic resonance: preliminary diagnostic accuracy," J Cardiovasc Magn Reson, vol. 15, p. 26, 2013.

[34] D. Chen, B. Sharif, R. Dharmakumar, L. E. Thomson, C. N. Bairey Merz, et al., "Quantification of myocardial blood flow using non-ECG-triggered MR imaging," Magn Reson Med, 2014.

[35] D. Likhite, G. Adluru, N. Hu, C. McGann, and E. V. R. DiBella, "Quantification of myocardial perfusion with self-gated cardiovascular magnetic resonance," J Cardiovasc Magn Reson, vol. 17:14, 2015.

[36] E. V. R. DiBella, L. Chen, M. C. Schabel, G. Adluru, and C. J. McGann, "Myocardial perfusion acquisition without magnetization preparation or gating.," Magn Reson Med, vol. 67, pp. 609-13, 2012.

[37] A. G. Christodoulou, T. K. Hitchens, Y. L. Wu, C. Ho, and Z. P. Liang, "Improved subspace estimation for low-rank model-based accelerated cardiac imaging," IEEE Trans Biomed Eng, vol. 61, pp. 2451-7, 2014.

[38] S. Giri, H. Xue, A. Maiseyeu, R. Kroeker, S. Rajagopalan, et al., "Steady-state first-pass perfusion (SSFPP): a new approach to 3D first-pass myocardial perfusion imaging," Magn Reson Med, vol. 71, pp. 133-44, 2014.

[39] B. Sharif, R. Dharmakumar, R. Arsanjani, L. Thomson, C. N. Bairey Merz, et al., "Non-ECG-gated myocardial perfusion MRI using continuous magnetization-driven radial sampling," Magn Reson Med, vol. 72, pp. 1620-8, 2014.

[40] M. Ishida, A. Schuster, G. Morton, A. Chiribiri, S. Hussain, et al., "Development of a universal dual-bolus injection scheme for the quantitative assessment of myocardial perfusion cardiovascular magnetic resonance," J Cardiovasc Magn Reson, vol. 13, p. 28, 2011.

[41] J. U. Fluckiger, M. C. Schabel, and E. V. DiBella, "Constrained estimation of the arterial input function for myocardial perfusion cardiovascular magnetic resonance," Magn Reson Med, vol. 66, pp. 419-427, 2011.

[42] D. Kim and L. Axel, "Multislice, dual-imaging sequence for increasing the dynamic range of the contrast-enhanced blood signal and CNR of myocardial enhancement at 3T," Journal of Magnetic Resonance Imaging, vol. 23, pp. 81-86, 2006.

Figures

Fig. 1: TurboFLASH acquisition of a contrast-filled stationary dog heart on the 3T scanner. The readout (vertical) and phase-encoding (horizontal) acquisition pixel dimensions are given on the images. Note that the dark rim artifact from strong edges perpendicular to the phase-encode direction is seen in a, decreases in b, and decreases further in c. From [4].



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)