Feasibility of contrast-enhanced coronary artery magnetic resonance angiography using compressed sensing

Cardiovascular magnetic resonance (CMR) is capable of comprehensive assessment of the heart including anatomy and function, myocardial tissue characteristics, and coronary artery without radiation exposure [1‐4]. Coronary magnetic resonance angiography (CMRA) has been developed over the past three decades as a possible noninvasive alternative for visualizing coronary arteries [5]. The balanced steady-state-free-precession (bSSFP) sequence has achieved great success in CMRA imaging at 1.5 T because of its intrinsic high blood signal intensity and blood-myocardial contrast [6]. Contrast-enhanced whole-heart CMRA with segmented fast low-angle shot (FLASH) 3D spoiled gradient-echo sequence at 3 T also demonstrated better depiction of coronary segments compared with bSSFP coronary CMRA at 1.5 T [7]. The shortening of CMRA acquisition time has been made possible by novel multi-channel cardiac coils and high parallel imaging factors [8]. Nevertheless, a CMRA scan still requires a comparatively long acquisition time, and is affected by the issues of operator dependency and limited ease-of-use, which limit its general applicability in clinical practice.

The recent development of the mathematical theory of compressed sensing (CS) has also been applied to CMR [9]. CS is based on reconstructing an image from an incompletely filled k-space [10, 11]. The CS technique is known to reduce scan time considerably, and was recently applied to cine CMR [12‐15]. Nakamura et al. reported that non-contrast CS CMRA could significantly shorten acquisition time, compared with conventional CMRA, in healthy subjects. In that study, almost all the CMRA scans were performed within 5 min [16].

In the clinical comprehensive CMR protocol, including cine CMR, perfusion CMR, late gadolinium enhancement (LGE) CMR, and CMRA, there is a waiting time of about 10 min after the injection of contrast medium for LGE CMR, and the CMRA scan is usually performed at the end of the protocol. We hypothesized that CMRA using CS could be routinely obtained within the LGE waiting time, thereby decreasing the total examination time and reducing patient burden. Therefore, the aim of this study was to verify the feasibility of CS CMRA scanning during the LGE waiting time in a clinical protocol.

All CS and conventional CMRA scans were successfully performed for all 50 patients. The LCX proximal and distal segments (#11, #13) of one patient showed hypoplasia and were excluded from the analysis. In total, 149 vessels and 348 coronary artery segments from 50 patients were evaluated. The median total scan time, including navigator efficiency, was 207 s (163, 259 s) for CS and 785 s (698, 975 s) for conventional CMRA (p < 0.001). The respiratory acceptance rate was 49.7 ± 9.5% for CS CMRA, and 51.4 ± 8.7% for conventional CMRA (p = 0.077). The heart rate was 66.5 ± 11.4 bpm for CS CMRA and 67.2 ± 12.8 bpm for conventional CMRA (p = 0.79). CS reconstruction was performed inline at the scanner at the end of the acquisition using graphical processing units (GPUs) on the MR reconstruction system. We could reconstruct the CS CMRA images in approximately 2 min.

From this prospective, two-center study, we make the first report on the feasibility of contrast-enhanced CS CMRA in a clinical protocol. CS CMRA enabled a substantial reduction in acquisition time while maintaining image quality compared with conventional CMRA. Due to the short acquisition time, all CS CMRA scans were successfully performed within the waiting time (10 min) for LGE in the clinical protocol.

CMRA has emerged as a method capable of providing visualization of coronary arteries without ionizing radiation [27, 28]. Shorter acquisition times have become possible using bSSFP sequences, parallel imaging with multi-channel coils, respiratory gating with navigator echoes, and improved strategies for k-space sampling [29, 30]. Today, the major problems with CMRA are the relatively long acquisition time and image artifacts caused by motion instability during the scan. Recently, new rapid imaging methods such as self-navigation and CS have emerged which can overcome these problems [18, 31‐33].

In our study, CS CMRA considerably shortened the acquisition time compared with conventional CMRA in clinical patients. In the clinical protocol, LGE CMR is commonly obtained 10 min after injection of contrast medium [34]. Our results showed that the examination time of CS CMRA, including preparatory steps such as the detection of the data acquisition window and the administration of nitroglycerin, is short enough to allow the scan to be performed within the waiting time between contrast injection and LGE scan.

The acquisition time of CS CMRA was very short and the acquisition data was smaller than conventional CMRA; therefore, the image quality of CS CMRA might be expected to be poorer. While a previous 3 T non-contrast study [16] showed inferior results from CS CMRA in terms of image quality and visible vessel length compared with conventional CMRA, our study did not reveal significant differences in either aspect. Furthermore, we found no significant difference in the quantitative evaluation of image quality as defined by sharpness. Several reasons were considered as to why the qualitative and quantitative evaluations with CS and conventional CMRA were not significantly different. Since conventional CMRA is performed at the end of the examination, it may be influenced by the patient’s movement due to fatigue from the long CMR examination. In our study, CS CMRA could be performed at an early stage, thus decreasing patient fatigue and movement. The motion of the heart through the cardiac and respiratory cycles can also cause blurring and artifacts in CMR studies [35]. The short acquisition time of CS CMRA may also reduce the artifacts produced by unstable respiratory conditions and motion during the scanning (Fig. 4). Respiratory self-navigation has been proposed as an alternative approach to compensate for respiratory motion while also increasing scan efficiency [18, 31‐33]. Further shortening of the imaging time can be expected by the combination of CS and self-navigation.

Differences in contrast agent concentration in the coronary artery at the time of imaging should also be considered. Contrast medium improves the signal-to-noise and contrast-to-noise ratios in CMRA, with an emphasis on the role of T1-shortening [36]. In our protocol, CS CMRA scans were performed immediately after injection of the contrast medium when the concentration of contrast agent was high. In comparison, since conventional CMRA was performed after LGE, the contrast agent had washed out over time and its concentration was relatively lower. Double-dose contrast medium (0.2 mmol/kg) is widely used for LGE [37]. However, in our study, single-dose (0.1 mmol/kg) medium was used because double-dose is not permitted in Japan. The single-dose was likely to reduce its concentration over time faster than the double-dose, leading to an advantage for CS CMRA that could be performed immediately after the injection of the contrast medium. In the CS technique, the contrast in the image plays a major part in the ability to reconstruct vastly undersampled images. High contrast often results in large distinct sparse coefficients [9]. These reasons may help to explain the lack of significant differences between CS and conventional CMRA for qualitative and quantitative evaluations. Contrast-enhanced CS CMRA could shorten the acquisition time considerably while maintaining image quality compared with conventional CMRA.

In this study, we verified that adequate CS CMRA could be performed during the waiting time in a normal clinical protocol. Preparatory steps such as the detection of the data acquisition window and the administration of nitroglycerin were performed after contrast medium administration so as not to disturb the clinical examination. If preparatory steps are performed before contrast medium injection, it can maximize the blood gadolinium concentration during CS CMRA acquisition and allow more time for CS CMRA scanning.

In addition, CS can also potentially improve spatial resolution [38]. In our results, CS CMRA scan times were significantly shorter than the 10-min limit imposed by the LGE waiting time. Therefore, by spending more time on the CS CMRA acquisition, image quality can be improved, or isotropic resolution can be achieved.

In the future, further high resolution CMRA imaging may allow for detailed coronary artery assessment. We can expect such new protocols to replace conventional CMRA with CS CMRA performed during the waiting time. This protocol will significantly shorten the total CMR examination time and lead to a reduction in patient burden.

The acquisition time of CS CMRA is short enough to allow the scan to be performed within the waiting time for LGE, thereby shortening the protocol time while maintaining image quality as compared with conventional CMRA performed after LGE imaging. These findings suggest the possibility of a faster CMR protocol.