Cardiac gating calibration by the Septal Scout for magnetic resonance coronary angiography

Magnetic resonance coronary angiography (MRCA) is a potential diagnostic tool for coronary artery disease (CAD). Compared to the current gold standard, x-ray angiography, benefits of MRCA include three-dimensional visualization of coronary vessel lumens, and not subjecting patients to the risks associated with catheterization [1, 2] and ionizing radiation [3]. MRCA, however, requires long acquisition times that span multiple heartbeats. Cardiac motion is particularly problematic for MRCA because high spatial resolution is required for diagnosing CAD.

Prospective cardiac gating remains the most effective tool for reducing cardiac motion artifacts in MRCA. The general principle behind cardiac gating is to synchronize, for each heartbeat, the imaging window with the diastasis period. Shechter et al. and Johnson et al. have measured coronary artery velocities during diastasis to be on the order of 10 mm/s [4, 5]. Several authors have also estimated imaging window durations within diastasis that will limit blur artifacts to within sub-millimetre pixels. Their estimates include: 66 to 330 ms with a mean of 187 ms for the entire coronary tree [4]; 66 to 220 ms with a mean of 120 ms for the right coronary artery (RCA) [6]; and, 65 ± 42 ms for the mid-RCA [7].

Cardiac gating is most commonly facilitated by the use of (1) the electrocardiogram (ECG) for monitoring ventricular systole onset, and (2) a cine cardiovascular MR (CMR) video of a heartbeat obtained prior to the MRCA scan for finding the timing of diastasis relative to the ECG. Yet, the use of cine-CMR to determine ventricular diastasis may produce cardiac gating errors due to insufficient temporal resolution of the cine-CMR method.

The temporal resolution of a conventional cine acquisition depends on the number of k-space lines acquired per segment (of k-space) per heart beat, herein denoted lines per segment (LPS), multiplied by the acquisition time of each k- space line, commonly known as the repetition time (TR). The number of segments used to complete k-space coverage dictates the number of heartbeats required for the total scan duration. For example, consider the cine parameters: heartbeat duration, 960 ms; field of view (FOV), 35 cm; number of frames, 30; TR, 4 ms; partial Fourier acquisition matrix size, 160 × 256; and LPS, 16. The cine acquisition would span 10 heartbeats, have a true temporal resolution of 64 ms. Extending this example with approximate motion parameters, a mid right coronary artery moving at a peak velocity of 12 cm/s during rapid filling and decelerating constantly over 100 ms toward rest during diastasis will traverse 1.5 mm in the 50 ms preceding diastasis. The reverse motion may be observed transitioning from diastasis to atrial contraction. Therefore, during transitional cardiac phases pertinent to ECG gating, the conventional cine acquisition may not provide the sufficient frame rate to resolve the motion of an RCA segment. In addition to inadequate temporal resolution, cine acquisitions are susceptible to temporal data-mixing if the subject’s heart rate varies during the scan duration.

In this paper, we present a new CMR technique, the Septal Scout, for determining the timing of the diastasis period. The Septal Scout measures ventricular septal motion. We demonstrate that cardiac gating windows determined by the Septal Scout produce sharper MRCA images than windows determined by cine-CMR. To facilitate a user-independent assessment of the diastasis period from cine-CMR data, an empirical approach was chosen based on identifying a plateau period of high frame-to-frame correlations during diastole [8].

In this study, we have successfully applied the MR Septal Scout technique to find more accurate cardiac gating windows than those obtained by cine-CMR. The use of gating windows determined by the Septal Scout led to a significant increase in vessel sharpness in MRCA images.

The principle behind the Septal Scout is to trade off extraneous spatial data, specifically a dimension of spatial encoding, for faster sampling of a targeted component of cardiac motion, specifically, the 1D long-axis motion of the basal ventricular septum. While the comparison of the Septal Scout to the cine-CMR technique is affected by other factors such as (1) imaging tissues in the Septal Scout plane which is perpendicular to the 4-chamber plane, and (2) acquiring 1D vs 2D data, it is believed that the superior temporal resolution of the Septal Scout is the most important factor for its improved performance.

In the subcohort comparison between the Septal Scout and TDE methods, the timing differences of the estimated diastasis periods sometimes reached as high as 15–20 ms. Since the Bland-Altman plots (see Figure 2) reveal no biases, this may reflect inherent variability within the methods.

The Septal Scout technique was superior to the cine-CMR technique in terms of being able to adapt to the HRV that was observed during a breath hold. For cine-CMR, modest HRV may cause incorrect data binning, which in turn may cause motion blurring. In contrast, the Septal Scout essentially freezes cardiac motion within its 10-ms frames. The superior acquisition speed permits the identification of the diastasis window on a per heartbeat basis, and provides a set of gating windows over multiple heartbeats that varies with heart rate. This ability to determine a gating window that is robust across an observed range of HRV during a breath hold is believed to have contributed to the ability of the Septal Scout method to obtain sharper MRCA images in this study.

As a result of the slice excitation, the Scout Plane extends beyond the beating septum to include many unwanted static signal sources such as from the chest and back. The Septal Scout acquisition may alternatively be obtained by the use of 2D RF excitation pulses to further isolate the ventricular septum. For example, a 2D spiral RF pulse [11, 12] may be used to excite a cylinder of tissue along the septal wall on the 4-chamber long-axis plane. Note that 2D RF pulses typically require longer durations. This alternative implementation must therefore consider the trade-off between the duration of the excitation pulse, the quality of the excitation profile, and the resultant TR. Generally, sharper excitation profiles require longer excitation pulses. A TR equal to or less than 10 ms should be maintained to provide sufficient temporal resolution for motion tracking. The design specifics of this alternative excitation method are left for future investigation.

The use of a 10-ms TR for the Septal Scout may introduce susceptibility-related artifacts that otherwise may be avoided with a shorter TR. In this study, however, this did not appear to have affected the ability of the Septal Scout to monitor the displacement of the basal septum and identify periods of quiescence. This is likely because the basal septum is a small ROI in a well-shimmed area.

The FWHM assessment of vessel profiles required the reformatting of vessel segments into a cross-sectional orientation. Theoretically, the asymmetric intrinsic resolution of the MRA data (1.5 × 1.5 mm² in-slice; 2 mm through slice) may affect the FWHM comparison if the actual vessel orientations were significantly different between the cine-guided vs. the septal scout-guided acquisitions. For example, if one method produced a gating window during which a vessel segment is oriented much more perpendicular to the slice orientation of the 3D image acquisition, the vessel cross-sections would appear sharper due to the higher in-slice resolution. We did not, however, observe this effect. The intrinsic resolution was near isotropic. And, this effect should not have systematically favoured one method of the study in particular.

The Septal Scout, in general, is sensitive to placement. There are two main types of septal localization errors: vertical and horizontal. Prescribed from a 4-chamber long-axis cine, vertical localization is achieved by aligning the Scout Plane such that it contains the septal wall throughout the cardiac cycle. Missing the septum at any time will result in the loss of the signal of interest. This type of error should be avoided. The prescription of the 4-chamber long-axis plane is the horizontal localization. Deviation by the 4-chamber cine from the horizontal plane of the heart is, however, not a significant concern. This is because the long-axis septal motion will still be the dominant component of motion observed on the Septal Scout. Assuming a good placement, respiratory and general patient motion may still introduce septal localization errors. Currently, this method relies on the use of breath holds to suppress respiratory motion. Like previous authors, this was found to work well for breath holds near or under 20 seconds [13].

In this study, a gradient optical flow calculation [10] was used to determine septal motion from the Septal Scout data. A correlation-based alternative was considered during experimental design using preliminary data. In this approach, a line ROI from a Septal Scout projection was tracked using maximum correlation to a new position in the subsequent projection. The ROI is updated at the new position, and tracked in the subsequent projection, and so on. In our initial experience, the correlation approach was inferior to the optical flow approach because the line ROIs have very few data points to statistically overcome image noise.

In CMR, navigators are 1D signals that monitor the displacement of an edge structure such as the diaphragm during tidal breathing [14]. Navigators are typically performed continuously (on their own and interleaved with imaging) and processed in realtime in conjunction with ECG gating to provide triggers for starting and stopping image data acquisition depending on the navigator position and cardiac phase. The Septal Scout may potentially be adapted into a Septal Navigator. The Septal Navigator function would be a realtime cardiac motion signature; it also may be compatible with a respiratory navigator. In an integrated gating system, the ECG R-peak may be used to trigger contrast preparation and the Septal Navigator may be used to trigger the start and end of imaging per heartbeat. A possible limitation may arise from the Septal Navigator interfering with the image acquisition during diastasis. The Scout Plane excitation may saturate signal sources in coplanar segments of coronary arteries, making them unavailable for imaging. 2D RF pulse excitations may be a solution being less intrusive than a slice excitation. Another issue is that the Septal Navigator may not maintain steady state magnetization, and may require a switch to low-flip-angle spoiled gradient recall methods instead of SSFP. These potential considerations are left for future investigations.

Direct cardiac motion monitoring is a shared concept between the Septal Scout and the anterior left ventricular (LV) wall navigator presented in the past by Stuber et al. [15]. A major distinguishing feature, however, is the location of the Septal Scout versus that of the LV navigator. Different epicardial regions have been known to have different diastasis timings [4, 5, 16]. The Septal Scout is more suitable for whole-heart angiography; basal septal diastasis is an accurate surrogate for ventricular diastasis [9]. To our knowledge, there has not been work presented on the CMR monitoring of septal motion for determining cardiac gating windows.

The performance of the Septal Scout in the presence of cardiac pathology has yet to be tested. In general, septal wall defects will likely change the relationship between septal and ventricular diastasis. One particularly problematic condition is known as the Septal Bounce, which is a paradoxical bouncing motion of the IVS initially toward and subsequently away from the LV during the beginning of diastole [17]. This condition is typically observed in constrictive cardiomyopathy, and is visible on a 4-chamber long-axis cine CMR. Septal Bounce may cause vertical plane localization errors in the Septal Scout technique. The behaviour of the Septal Scout in the presence of various septal defects should be an investigative component in future patient studies. It is important to know if and when septal diastasis is no longer an accurate surrogate for ventricular diastasis.

In our future work, we intend to explore the use of the Septal Scout to determine the onset of ventricular systole, which is currently detected by the R-peak of the ECG. This provides the benefit of not having to maintain an ECG signal to perform cardiac-gated MR imaging. Currently, the ECG signal may arbitrarily deteriorate due to loosened connections at the chest electrodes; also, R-peak detection may fail due to significant T-wave amplification [18], which may sometimes be a problem even with the use of the vector cardiogram.

In this chapter, the MR Septal Scout was shown in a healthy volunteer study to be effective at determining accurate cardiac gating windows. Specifically, this new technique is consistently more accurate than the cine-CMR method, the current clinical standard. The improvement in cardiac gating accuracy provided by the Septal Scout lead to sharper coronary MRA images. Furthermore, the Septal Scout was shown to provide diastasis windows that are in close agreement with those provided by TDE measurements of septal motion.