Phase-contrast MRI is an established, routinely applied technique for non-invasive, clinical assessment of cardiovascular physiology. Most implementations of PC-MRI employ segmented k-space techniques to shorten imaging times in order to permit breath-held imaging, which in turn minimize respiratory artifacts. However, this is done at the expense of temporal resolution. SENSE can improve scan time, temporal resolution, spatial resolution, or a combination of these factors. Increasing temporal resolution improves clarity of rapidly moving parts, such as the heart valves. Increasing spatial resolution helps imaging small structures and reduces partial volume effects. Decreasing scan time also helps improve patient compliance. Therefore, the incorporation of SENSE with PC-MRI is a logical and powerful step. Despite demonstration of feasibility [11
], a clinically practical and efficient implementation has yet to be reported. While the accuracy of SENSE PC-MRI has been assessed in calculating shunt ratio [11
], it has yet to be evaluated in velocity measurements, which are needed for pressure gradient estimation.
In this study, we explored the approach of using a single, low resolution, low contrast, high-SNR, 3D-SPGR sequence to acquire the calibration scan for all subsequent SENSE-PC scans. For the case of shunt ratio quantification using conventional PC-MRI, the total scan time would include the time for two scans of the aorta and pulmonary artery. If a non-accelerated PC-MRI sequence takes 20 seconds per slice, then the total scan time for two slices would be about 40 seconds. In our approach, the single 3D calibration scan takes about 10 seconds and the total scan time is 30 seconds, close to a 25% savings. In clinical practice, the savings in scan time is likely greater since additional flow planes are routinely acquired. Alternatively, the savings in scan time can be traded for improved temporal resolution, spatial resolution, or both. To maintain the focus of this study, we did not explore these additional possibilities.
In our volunteer study, we showed that the performance of our SENSE-PC implementation is generally compatible with an established, non-accelerated PC-MRI sequence. The direct comparison between SENSE-PCx1 and Fastcard-PC of blood flow values showed a mean difference of 0.06 L/min, which is insignificant compared to a normal cardiac output of 5 L/min or a peak systolic aortic flow of 10 L/min. Similarly, the mean difference for peak velocity was 0.0053 m/s, which is insignificant compared to a normal aortic valvular velocity of 1–2 m/s.
Our repeatability study showed that consistency from scan-to-scan worsens with SENSE acceleration. This is reflected by an increased limit of agreements from 0.98 L/min at R=1 to 1.16 L/min at R=3 for flow, and from 0.11 m/s at R=1 to 0.17 m/s at R=3. These changes can be expected from the increased noise and decreased signal-to-noise ratio in SENSE accelerated images.
In the patient study, there is good agreement in flow and peak velocity for a reduction factors up to 3. A small positive bias, up to 0.12 L/min, was detected for flow and a small negative bias, up to 0.083 m/s, was detected for peak velocity, but both are small compared to the normal physiological ranges. Beerbaum [11
] reported close agreement between SENSE-accelerated PC-MRI compared with non-accelerated PC-MRI for aortic flow, pulmonary flow, and shunt ratio. They employed a log-transformed Bland-Altman method to compare the ratio of two techniques and reported a 3% mean difference and a 24% maximum limit of agreement. In our study, we believe that a comparison of absolute difference is more appropriate since there is no expectation that error increases with the magnitude of the velocity or flow data. While our results qualitatively agree with their assessments concerning flow, a quantitative comparison of results was not possible.
We note in our study that accuracy decreases in pulmonary trunk with an increase in the reduction factor. However, this trend is not as prominent in the aorta or the coarctation. This phenomenon cannot be explained by g-factor difference since the g-factor is greater at the aorta as compared to the pulmonary trunk. One possible explanation may be the difference in orientation of the image planes for the aorta and the pulmonary trunk. For the aorta, the transverse image lies mostly in the axial plane, where the anterior-posterior dimension of the chest is small. In contrast, the transverse image for the pulmonary trunk lies mostly in the coronal plane, where the body fills the image. When SENSE undersamples k-space, the wrapping of mostly empty space in the axial plane produces less artifacts compared with scanning in the coronal plane. At higher reduction factors, separating aliasing artifacts with SENSE in the axial plane produces less error and artifacts than in the coronal plane.
Despite its benefits, SENSE PC-MRI introduces several problems not encountered in conventional PC-MRI. The fidelity of SENSE reconstruction is strongly dependent on the appropriate coil design and the proper match between the phase-array coils and the body size of the subject. The primary 8-channel cardiac coil that was used for this study is designed for adult body size and is not optimized for small pediatric patients, especially in the left-right direction. Errors in coil sensitivity maps in the chest wall, where fat signals are strong, may additionally contribute to inaccuracies.
One complication regarding SENSE PC-MRI is aliasing in full field-of-view image reconstructions [18
]. When the coil sensitivity maps do not fully correct for aliasing in the full field-of-view, artifacts are produced. In this study, we carefully avoided the prescription of aliasing in full field-of-view as much as possible. For those patients where we could not avoid this complication, we did not encounter appreciable changes in the flow and peak velocity measurements.
For estimating peak velocities, the SENSE PC-MRI measurements introduce a slight negative bias. One possible explanation is that when peak blood flow velocities approach the maximum encoded velocity, additive noise beyond the maximum velocity would be registered as an aliased, negative velocity. This would introduce a negative bias to the estimation. As SNR decreases with greater reduction factor, this negative bias may also become greater.
Finally, our approach of scanning a calibration scan once, although efficient, may introduce misregistration artifacts when the patient moves or breathes differently over the course of the study. When this type of artifact becomes noticeable, the calibration scan must be repeated.
Some of the variations observed in this study may be caused by physiological changes. Beerbaum [19
] reported 5.3% variability in shunt ratio quantification in scans performed 3 times with pediatric populations. Powell [20
] reported ± 0.2 limits of agreement for normal volunteers with expected shunt ratios of 1.0, corresponding to 20% variability. It is expected that random patient movement or physiological changes occur which may affect flow measurements seen in this study.
In summary, a practical and efficient SENSE PC-MRI sequence has been implemented and tested in patients under clinical conditions. Flow and velocity data derived from SENSE PC-MRI were found compatible with conventional PC-MRI in the imaging of the great arteries. The efficiency gained from SENSE may be traded for shorter scan time, increased temporal resolution, increased spatial resolution, or combinations thereof. Thus, SENSE is a valuable and powerful tool that enhances the performance of clinical PC-MRI.