Rapid and Continuous Imaging for Body Magnetic Resonance Examinations
Abstract： Body MR Examinations would significantly benefit from a new paradigm that enables rapid and continuous imaging.
Need for Speed in Body MR Examinations
Noninvasive and robust assessment of human body structure and function has long been a target for a wide range of imaging modalities. Among those existing modalities, computed tomography (CT) has gained widespread attention for body exams due to its fast imaging speed and simple exam procedure, in which high resolution volumetric information can be obtained in a few seconds. However, CT has access to only a limited range of contrast mechanisms and perhaps more importantly, CT examinations must be associated with ionizing radiation, which has raised major concerns about the risk due to cumulated x-ray exposures.
Magnetic Resonance Imaging (MRI) is another promising modality for body exams. MRI is free of ionizing radiation, offering excellent soft-tissue characterization with high resolution and flexible contrast parameters, all of which allow visualization of both anatomical structure and functional information. However, although MRI has been increasingly used for Neuro and Musculoskeletal exams, its applications for body imaging, such as cardiac and abdominal assessment, are relatively limited. This barrier largely results from the stringent constraints on the MR imaging speed, which has led to relatively long examination times, and/or limited resolution and coverage. Moreover, due to the speed constraint, data acquisitions in MRI are often planned in oblique planes adjusted to the target anatomy, which reflects the underlying complexity for clinical utility of body MRI.
The needs for speed in body MRI are multifold. First, patients may not be able to hold their breaths for a long time during the scans. Thus, long examination times tend to increase patient’s discomfort, leading to unreliable results. Second, fast acquisitions are needed to catch moving structures (e.g., a beating heart or a moving liver) with minimized effects due to motion. In a dynamic contrast-enhanced (DCE) exam, fast imaging speed is also required to capture the quick passage of contrast agents, which are often injected to highlight particular internal structures and disease information. Finally, improved imaging efficiency could enable acquisition of an increased amount of information per unit time, which enhances the value of imaging studies and also reduces the cost of each scan.
Rapid Body MR Examinations
Since the introduction of MRI, tremendous effort has been devoted to improve imaging speed. The speed at which MR images can be acquired now has already been dramatically increased, with a combination of advances in MR hardware (e.g., better gradient system and coil design) and innovations in both imaging acquisition and reconstruction strategies.
The advent of multi-detector systems (arrays of radiofrequency coils for MR signal reception, also known as parallel imaging) [1-2] in the late 1990s enabled further improvement in MR imaging speed beyond previous limits by acquiring more than one measurements simultaneously. Since then, parallel imaging has motivated a great number of applications, especially in body MRI, which were previously challenging due to limited imaging speed. However, the acceleration capability of parallel imaging is fundamentally limited by the noise amplification, which increases non-linearly with increasing acceleration factor [2-3]. This fact posed a practical limit for further increase in imaging speed with parallel imaging alone.
The idea of compressed sensing [4-6], proposed in the 2000s, represents another powerful approach for increasing imaging speed in MRI by exploiting the compressibility or sparsity of images. Compressed sensing takes advantage of the fact that any image can have a sparse representation in some appropriate basis and thus it is possible to reconstruct the sparse information from incoherently undersampled measurements without loss of important information. Compressed sensing is especially useful for accelerating dynamic imaging, where temporal dimensions usually present significantly higher level of sparsity. Successful applications of compressed sensing have been demonstrated for cardiac MRI, abdominal MRI, MR parameter mapping, dynamic contrast-enhanced MRI, and many others. Moreover, appropriate combinations of compressed sensing and parallel imaging has also been proposed for additional improvement in imaging speed as compared with either of them alone [7-9]. After the introduction of compressed sensing to MRI, the enthusiasm for rapid MRI has been raised to a new level, and it is bringing rapid MRI into a new era of sparsity.
Rapid and Continuous Imaging for Body MR Examinations
Despite the large variety of rapid MRI techniques and remarkable advances in MRI speed over the last two decades, the paradigm of routine clinical settings still remains complex, given the diversity of acquisition choices and the adjustments needed to avoid or reduce the influence of unwanted effects, such as respiratory motion, cardiac beating, and others. This underlying complexity results from the fact that MR imaging protocols are traditionally designed as an ordered series of snapshots, which remains even with an order-of-magnitude increase in imaging speed. Although these simple and repeatable protocols are often employed in the conventional setting where ordered acquisitions are performed, a non-repeating acquisition is more preferred for compressed sensing MRI, since the increased dimensionality and information in the resulting datasets leads to higher correlation and incoherence that can be exploited in image reconstruction to enable better imaging performance. This fact, actually, starts suggesting a hint and tendency for a continuous data acquisition scheme in the era of sparsity to allow acquisitions of diverse and comprehensive image information in a single, rapid and continuous stream.
Examples of Continuous Body MR Imaging
In this section, two concrete examples, including one abdominal DCE-MRI example and one cardiac MRI example, are described for the applications of continuous body MR scans. In routine DCE-MRI exams of the abdomen (e.g., a liver), volumetric images at different contrast enhanced phases, such as pre-contrast phase, arterial phase and portal venous phase, are usually acquired sequentially during multiple breath-holds, as shown in Figure. 1a. Since contrast agent can be injected only once, images contaminated by motion-artifacts (e.g., ghosting or blurring) have to be used for compromised clinical diagnosis in the case of a failed breath hold. The GRASP (Golden-angle RAdial Sparse Parallel MRI) technique, as presented in , represents a better way of performing DCE-MRI of the abdomen during free breathing with a push-button continuous scan. The combination of continuous data acquisition using golden-angle radial sampling , compressed sensing and parallel imaging employed in GRASP enables flexible reconstructions that can be tailored for various clinical needs. As shown in Figure. 1b, different contrast enhanced phases can be retrospectively selected from the reconstructed image series. XD-GRASP (eXtra-Dimensional GRASP), as described in , further extended GRASP by reconstructing an additional respiratory motion dimension using the compressed sensing techniques, enabling one to handle respiratory motion without performing any explicit motion correction or compensation and without any assumption of motion model (e.g., displacement of affine transformation) during the image reconstruction process.
The idea of continuous MRI has also been applied for cardiac MRI. Promising results have recently been reported by different research groups to demonstrate the feasibility of a push-button free-breathing whole-heart MRI with simultaneous visualization of myocardial function and coronary artery anatomy at any cardiac phases [13-14]. Respiratory motion can be retrospectively corrected to enable nearly 100% scan efficiency. Golden-angle radial sampling employed in these studies ensures continuous data acquisitions without pre-defining temporal frames and there is little chance of missing key information with such imaging paradigm. The framework in one of the works has been further extended into a five-dimensional cardiac motion- and respiratory motion-resolved whole-heart MRI paradigm using XD-GRASP , where explicit respiratory motion correction is not needed anymore and images can be obtained at different cardiac and respiratory phases, with additional physiological information available along the respiratory dimension for potential clinical use. These works all have the potential to shift the cardiac MRI exam from a cumbersome workflow into a simple rapid and continuous procedure, as shown in Figure. 2.
Vision of Future Body MRI Paradigm
We are now in the era of sparsity, and the advent of compressed sensing has started changing the way we think about image formation. Sparsity offers an opportunity to connect rapid imaging with a simple and continuous data acquisition scheme for enhanced information content. Such a goal entails more than just acceleration of existing imaging protocols and it could represent a shift of the day-to-day clinical workflow from conventional time-consuming and tailored acquisitions towards rapid and continuous volumetric acquisitions with flexible reconstructions that can be adapted retrospectively for different clinical or research needs. The MRI acquisitions in this way may become multifaceted with diverse information streaming in constantly, and the inherent flexibility of MRI would become how one could retrospectively capture clinically useful information from the continuously acquired measurements . Furthermore, despite lots of challenges, the newly introduced concept of MR Fingerprinting  might be brought in someday in the future to enable a further improved continuous and comprehensive body MR paradigm.
The author would like to gratefully acknowledge the generous support from the Center for Advanced Imaging Innovation and Research (www.cai2r.net) at New York University School of Medicine. The vision outlined in this article is summarized from the paper “Sodickson DK, SPIE Medical Imaging 94170G-94170G-14” （ref 16）. Readers who have interest in learning more can go to that paper for more information, images, and videos.
 Sodickson DK et al. MRM 1997;38(4):591-603.
 Pruessmann KP et al. MRM 1999;42(5):952-962.
 Sodickson DK et al. Med Phys 2001;28(8):1629-1643.
 Candes E et al. IEEE 2006;52:489-509.
 Donoho D et al. IEEE 2006;52:1289-1306.
 Lustig M et al. MRM 2007;58(6):1182-1195.
 Liang D et al. MRM 2009 Dec;62(6):1574-84.
 Otazo R et al. MRM 2010;64(3):767-776.
 Lustig M et al. MRM 2010;64(2):457-471.
 Feng L et al. MRM 2014;72(3):707-717.
 Feng L et al. MRM 2015 Mar 25. doi: 10.1002/mrm.25665. [Epub ahead of print].
 Winkelmann S et al. IEEE Trans Med Imaging. 2007 Jan;26(1):68-76.
 Pang J et al. MRM 2014 Nov;72(5):1208-17.
 Coppo S et al. MRM 2015 Nov;74(5):1306-16.
 Feng L et al. ISMRM 2015. p27.
 Sodickson DK, SPIE Medical Imaging 94170G-94170G-14.
 Ma D et al. Nature 2013 Mar 14;495(7440):187-92.
Li Feng received his PhD in Biomedical Imaging with a focus on rapid and continuous MRI from New York University School of Medicine in 2015. He is currently staying as a Postdoctoral Fellow at the same institute, continuing the research on developing novel techniques for rapid, continuous and comprehensive body MRI. Li Feng received the Society for Cardiovascular Magnetic Resonance (SCMR) Early Career Award in Basic Science in 2014 and was selected as a Junior Fellow of International Society for Magnetic Resonance in Medicine (ISMRM) in 2015.