Intensity of arterial structure acquired by Silent MRA estimates cerebral blood flow

Background Cerebral blood flow (CBF) and the morphology of the cerebral arteries are important for characterizing cerebrovascular disease. Silent magnetic resonance angiography (Silent MRA) is a MRA technique focusing on arterial structural delineation. This study was conducted to investigate the correlation between Silent MRA and CBF quantification, which has not yet been reported. Methods Both the Silent MRA and time-of-flight magnetic resonance angiography scans were applied in seventeen healthy participants to acquire the arterial structure and to find arterial intensities. Phase-contrast MRA (PC-MRA) was then used to perform the quantitative CBF measurement of 13 cerebral arteries. Due to different dataset baseline signal level of Silent MRA, the signal intensities of the selected 13 cerebral arteries were normalized to the selected ROIs of bilateral internal carotid arteries. The normalized signal intensities were used to determine the relationship between Silent MRA and CBF. Results The image intensity distribution of arterial regions generated by Silent MRA showed similar laminar shape as the phase distribution by PC-MRA (correlation coefficient > 0.62). Moreover, in both the results of individual and group-leveled analysis, the intensity value of arterial regions by Silent MRA showed positively correlation with the CBF by PC-MRA. The coefficient of determination (R2) of individual trends ranged from 0.242 to 0.956, and the R2 of group-leveled result was 0.550. Conclusions This study demonstrates that Silent MRA provides valuable CBF information despite arterial structure, rendering it a potential tool for screening for cerebrovascular disease. Supplementary Information The online version contains supplementary material available at 10.1186/s13244-021-01132-0.

on intracranial artery delineation and cerebral blood flow (CBF) have been developed and applied clinically [1].
Conventionally, three-dimensional time-of-flight magnetic resonance angiography (3D TOF-MRA) has been used for evaluating morphology of intracranial arteries [2]. This contrast-agent-free technique uses the flow of spins to generate vessel contrast; it is sometimes considered as a follow-up imaging alternative to digital subtraction angiography [3]. Unfortunately, spin saturation effects (in slow flow) and phase dispersion artifacts (in turbulent flow) decrease signal intensity in 3D TOF-MRA [4][5][6], leading to overestimation of the severity of intracranial artery disease. Additionally, 3D TOF-MRA is prone to magnetic susceptibility and radiofrequency shielding artifacts. In addition, cerebral flow rate cannot be determined using 3D TOF-MRA.
As a result, phase-contrast quantitative magnetic resonance angiography (PC-MRA) without contrast agent has been developed to measure CBF as a volume flow rate (ml/min) [7,8]. Zarrinkoob et al. [9] reported the distribution of total cerebral arterial flow across variations in age, sex, and anatomy using high-resolution PC-MRA establishing a normative reference value for blood flow in major cerebral arteries in healthy people [7]. Although PC-MRA can provide multidirectional flow and collateral flow, it does not improve stenosis detection in major intracranial vessels compared to 3D TOF-MRA because the optimized velocity-encoding gradient affects predominantly small vessels and the reduced number of partitions [10].
Silent magnetic resonance angiography (Silent MRA) is another MRA imaging technique [11][12][13]. As with 3D TOF-MRA, Silent MRA does not require contrast agent; it is a non-invasive perfusion imaging technique that uses continuous arterial spin labeling with a long radiofrequency inversion pulse, and it labels blood within the carotid arteries as an endogenous tracer [14,15]. Subtracted by another image dataset without the use of arterial spin labeling, Silent MRA can depict the intracranial arteries without background tissues. However, with arterial spin labeling technique, Silent MRA tends to have drop-off on inflow enhancement [16]. Previous study reported that Silent MRA showed low signal intensity in distal vessels because of poor inflow (two of 27 intracranial Silent Scans) [17]. On the other hand, it implied that the signal intensity obtained by Silent MRA contained not only the structure of arteries but also the flow information.
Moreover, with the use of a zero echo time technique, Silent MRA is able to minimize the phase dispersion of the labeled blood flow signal and decrease magnetic susceptibility compared to 3D TOF-MRA [18]. Its use has recently been proposed for assessing vascular lesions such as those in treated intracranial aneurysms [18][19][20][21] and in cerebral arteriovenous malformations [22,23]. These studies show that it provides excellent architectural visualization in coiled aneurysms and flow through intracranial stents.
Studies have focused primarily on delineating vascular lesion structures using a combined zero-echo-time and arterial spin labeling technique, but none have thoroughly investigated the use of Silent MRA for estimating CBF. Therefore, we aimed to explore the usefulness of Silent MRA for flow estimation. We studied the correlation between Silent MRA and CBF as estimated using PC-MRA in a healthy population.

Participants
Seventeen healthy volunteers (8 women; mean age, 33.8 ± 7.1 years) joined this study between October 2018 and July 2019. None had claustrophobia, psychological disorders, cardiac pacemakers, contraindications to magnetic resonance imaging, or metal implants, and none were pregnant. During the entire scanning session, the participants were to maintain a motionless head. Participants remained awake to prevent unwanted motion artifacts. This study was approved by the Research Ethics Committee of Taipei Medical University-Joint Institutional Review Board (N201803017), and informed consent was obtained from all participants.

MRA scanning
All magnetic resonance images were acquired using a 3.0-T clinical scanner (Discovery MR750w; GE Healthcare, Milwaukee, USA) equipped with a 24-channel Geometry Embracing Method head-and-neck coil for signal detection and a whole-body coil for radio-frequency excitation. Both the Silent MRA and 3D TOF-MRA scans were used to acquire the arterial structure before performing 2D PC-MRA. In the 2D PC-MRA flow-measuring scan, a Non-Invasive Optimal Vessel system (NOVA; VasSol, Inc., Chicago, IL, USA) was used to perform the quantitative flow measurements in 13 cerebral arteries: the basilar artery, the bilateral anterior cerebral arteries, the superior branches of the bilateral anterior cerebral arteries, the bilateral middle cerebral arteries, the bilateral posterior cerebral arteries, the bilateral vertebral arteries, and the bilateral internal carotid arteries. The parameters for acquiring the PC MRA were: repetition time/echo time, 12.82 ms/5.728 ms; flip angle, 25°; field of view, 16 cm; matrix, 256 × 256; section thickness, 5 mm; NEX, 1; bandwidth, about 162.734 kHz; VEC, automatically selected by NOVA software. To minimize the effects of turbulence, flow was measured at the point furthest from vessel turns within the designated segment. Perpendicular imaging for any given vessel was automatically created by the NOVA system so that arterial flow could be accurately measured. A laminar flow model was used to calculate vascular flow rates. The vessels used to designate flow status were those in distal territories of the vertebrobasilar tree. Quantitative flow was then measured in the chosen vessels via the 2D PC-MRA technique using the NOVA system. Data were acquired under the guidance of one experienced (25 years) radiological technologist.

Silent MRA data analysis
We compared the arterial structure and vessel uniformity of bilateral internal carotid arteries (ICAs) between Silent MRA and 3D TOF-MRA. To interpret the flow information obtained using Silent MRA, average image intensities based on the exact location of the 13 cerebral arteries in PC imaging were calculated, and the flow-encoding Silent MRA signal distribution was compared with that obtained using PC flow imaging. Because the baseline of each individual Silent MRA dataset resulted in a different signal level, the image intensities of the 13 arterial areas were normalized to the central regions of the selected ROIs located prior to the bilateral cervical segments of ICAs (Fig. 1). In individual analyses, the normalized intensities of the arterial regions, generated using Silent MRA, were correlated to the corresponding flow rates generated using PC imaging. In the group-level analysis, one linear regression was applied across all the normalized arterial intensities and their corresponding flow rates, allowing investigation of the relationship between the normalized arterial intensities and CBF. To explore the relationship between the normalized arterial intensity from the Silent MRA and the quantified CBF, linear regression with 95% confidence intervals was calculated. Statistical analyses were performed using R version 3.5.1 (R Core Team 2018. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL http:// www.R-proje ct. org/).

Results
Compared with 3D TOF-MRA, Silent MRA provided more detail in the arterial structure and a more uniform intensity in the petrous segments of internal carotid arteries (one example was listed in Fig. 2, and all the dataset were listed in the Additional file 1: Fig. S1), and it showed comparatively less signal loss in the Silent MRA. Additionally, its structural map showed gradient  Table 1 compares the CBF values generated using 2D PC-MRA for each of the 13 arteries to the image intensities generated using Silent MRA at the same arterial location. The normalized arterial intensity was adopted at each arterial location in the group-level analysis, and a linear relationship (R 2 = 0.550) was found (Fig. 4). For every participant, the normalized arterial intensities across the 13 arterial regions were highly positively correlated with the corresponding CBFs: R 2 ranged from 0.242 to 0.956. Linear regression was also used to determine the relationship between the normalized arterial intensities in the Silent MRAs and the quantified CBFs. With 95% confidence, the group-level result predicts the CBF range when using the following:

Discussion
The novel contributions of this study are its investigation into the flow-weighted image intensity of Silent MRA, and the determination of a relatively straightforward linear relationship that flexibly connects the normalized arterial intensities in Silent MRA with quantified CBF. Using linear regression and the 95% confidence interval, the normalized arterial intensity provided by Silent MRA can be indirectly used to predict the corresponding CBF range in these arterial regions. For example, a normalized arterial intensity of 0.7 in the Silent MRA image corresponds, with 95% confidence, to a CBF ranging from 192.1 to 286.9 ml/min. Therefore, one 5-min Silent MRA can provide both the structural information and a relative quantification of CBF from a 2D PC-MRA, implying its potential role in screening for arterial disease.
Silent MRA has two primary clinical disadvantages. There is marked background suppression with loss of anatomical landmarks, particularly with normal variants of intracranial arteries. However, background suppression shows better blood vessel contrast. Baseline image studies such as computed tomography angiograms or digital subtraction angiographies are performed first and can be used as references, rendering Silent MRA useful as follow-up imaging for treated intracranial arteries. Secondly, compared to 3D TOF-MRA, the longer acquisition time could increase the possibility of motion artifacts.

Limitation
This study has several limitations. First, it is a singlecenter cohort study with small number participants and fixed scanning parameters; however, it provides promising preliminary results not yet found in the literature. Second, it was limited to young healthy participants. Initially, this was part of the study design to avoid potential bias associated with disease states. Further prospective study with larger population is needed to validate the diagnostic value of Silent MRA in estimating cerebral Fig. 4 Linear correlations between the normalized arterial intensities in Silent MRA and CBF determined using 2D PC-MRA. a Correlation between the normalized arterial intensities and the corresponding CBFs was highly linear (R 2 = 0.550). The dark grey area indicates the 95% confidence interval of the linearly fitted line. b At the individual participant level, normalized arterial intensity was also highly linearly correlated blood flow in older participants with cardiovascular risk or intracranial stenosis. Third, Silent MRA is independent of flow direction. Therefore, cerebral hemodynamic changes with arterial steal and flow reversal, as with subclavian steal syndrome or reversed Robin Hood syndrome, might not be detected. This could be solved by inducing advanced arterial spin labeling such as vesselselective labeling [31,32]. Fourth, it is difficult to select a common standard from images by Silent MRA for normalization due to the absence of background tissues. The standard used for normalization in this study is the ROI intensity of bilateral ICAs. However, it also contains the flow information within the selected ROIs. This results in the vary correlation between the intensity of Silent MRA and the flow by 2D PC-MRA among participants, which responses to the lower group-leveled result (R 2 = 0.550) while compared to individual results (most R 2 > 0.7). It may be solved by a better standard, such as the nonspin-labeling images reconstructed from the individual raw k-space, used for normalization. Fifth, variation of the intracranial artery is not unusual. For example, some degree of asymmetry between the two A1 segments was noted in up to 80% of cases [33], and the frequency of hypoplastic vertebral artery has been reported in up to 26.5% of normal healthy population [34]. The distribution of the signal intensity and vessel contrast generated by Silent MRA may result in alterations. For the middle cerebral artery region, we only measured M1 segments bilaterally, and further labeling of the M4 segments should be evaluated in future study. Lastly, Silent MRA is currently not feasible for single artery measurement and does not provide assessment of its CBF quantification.
In conclusion, the normalized arterial intensity in Silent MRA and the corresponding CBF generated by 2D PC-MRA are positively correlated, and the former can be used to predict the range of the latter. By providing a better depiction of the arterial structure and additional CBF information compared to 3D TOF-MRA, Silent MRA is a potential tool for screening steno-occlusive disease and providing supporting imaging information.