Inuence of the Area of the Aqueduct on Quantication of Stroke Volume and Max Velocity in Healthy Volunteers Using Phase Contrast Cine MRI

Background: The relationship of the area of the aqueduct on quantication of the aqueductal stroke volume (SV) and max velocity need further investigation. Our aim was to assess the in ﬂ uence of the area of the aqueduct on quantication of the aqueductal SV and max velocity measured with phase contrast magnetic resonance imaging (PC-MRI) within the cerebral aqueduct at the level of the intercollicular sulcus. Materials and Methods: Nine healthy volunteers (mean age 29.6 yrs) were enrolled in the study and brain MRIs were performed on a 3.0T system. Quantitative analysis of aqueductal cerebrospinal uid (CSF) ow was performed using manual regions of interest (ROI) placement. ROIs were separately drawn for each of 12 phases of the cardiac cycle, and changes in aqueduct size during the cardiac cycle were determined. Stroke volumes were calculated uses the rst and ninth aqueductal ROIs and compared to each other. Max velocities at the 12 phases were also collected, and the relationship between the area and max velocity and the impact on SV were analyzed. Results: There was variation in the size of the aqueduct during the cardiac cycle, the rst area (cid:0) S1 (cid:0) was larger than the ninth (cid:0) S9 (cid:0) . The rst max velocity (cid:0) Vmax1 (cid:0) was less than the ninth (cid:0) Vmax9 (cid:0) . Additionally, there was a signicant different between the stroke volume calculated using the rst aqueductal ROI (SV1) and the ninth (cid:0) SV9 (cid:0) . Conclusions: There is variation in the size of the cerebral aqueduct which is used to calculate stroke volume and other CSF ow parameters during the cardiac cycle. The maximum velocity may be inversely proportional to the area of the aqueduct. In order to establish reliable reference values for CSF ow parameters in future studies, a variable ROI, to account for cardiac cycle variation, should be considered and incorporated.


Introduction
Phase contrast cine magnetic resonance imaging (PC-MRI) has been used to measure cerebrospinal uid (CSF) ow dynamics. Stroke volume is de ned as the net ow of CSF in a de ned region of interest (ROI), and is often measured in the cerebral aqueduct. [1][2][3][4][5] Previous studies have evaluated CSF ow in healthy subjects as well as pathological ow in hydrocephalus [1,6,7,8] , aqueduct obstructions [9] , Chiari malformation [10] , and interventions [11,12] . PC-MRI has also has been used to evaluate the patency of third ventriculostomies as well as shunt catheter ow [13][14][15] .
As the CSF ow parameters are calculated, they depend on different factors such as speci c MR machine, eld strength, sequence parameters, post processing software [4] , as well as patient age and gender [16] . Additionally, aqueduct area may change with age, disease and intracranial pressure (ICP) [17] , but its change has not been assessed with the cardiac cycle.
As we know, the blood ow velocity is inversely proportional to the cross-sectional area of the blood vessel. Karin Markenroth Bloch et al. [18] reported that the maximum velocity-time curve of CSF in the aqueduct was similar to the blood ow velocity curve in a cardiac cycle. As our study found that there was variation in the size of the aqueduct during the cardiac cycle, there is a hypothesis that the CSF ow velocity is inversely proportional to the size of aqueduct, so we try to prove it.
Furthermore, despite the development of several semiautomated segmentation methods [19,20] , manual ROI segmentation is still widely used [1][2][3][4][5][6][7][8][9][10][11][12][13][14][15][16][17][18] . Although quantitative assessments of stroke volume have been conducted at the aqueductal level, current calculation methods only select a static area of ROI, so there is limited data on the impact of changes in area of the aqueductal ROI on stroke volume. Therefore, the objective of this study was to determine how much the size of the aqueduct varied during the cardiac cycle and how ROI size affects the calculated stroke volume measured by PC-MRI.

Participants
The study protocol was approved by the local Ethics Committee and informed consent was obtained from all participants. Nine healthy volunteers were identi ed and enrolled in the study. Subjects with a cardiac arrhythmia were excluded from the study.

MR Acquisition
PC-MRI was performed using a 3.0-T MRI scanner (Phillips Achieva 3.0T TX, The Netherlands) with a 16channel head coil and MR Extended Workspace R2.6.3.5 station (Philips, The Netherlands). A routine clinical protocol was used to obtain T1-weighted (T1w), T2w and T2w FLAIR images. Product sequences, CSF-QF sequences were then used to analyze CSF ow. The following imaging parameters were used: minimum TR and TE, ip angle = 15 , FOV = 150 mm * 150 mm, matrix = 256 * 256, slice thickness = 4 mm; ow direction from feet to head; velocity encoding value was 12cm/s. The acquisition time was 3-6 min, based on the participant's heart rate. A single slice was acquired with plane orthogonal to the aqueduct at the level of the intercollicular sulcus (Fig. 1). Retrospective cardiac gating via peripheral pulse device, known as peripheral pulse triggering, was performed. The data was binned into 12 phases during a cardiac cycle, based on the R-R interval. CSF ow was quantitatively analyzed with software (CSF-QF). The velocity versus time curve should be nearly sinusoidal with a period equal to the R-R interval in the ECG. [5,14] Analysis We rst sought to measure the change in area of the aqueduct during a cardiac cycle. All 12 scans in a cardiac cycle were extracted, and in each image, the ROI was drawn by two attending radiologists, to enclose the aqueduct with as little surrounding tissue as possible ( Fig. 2A). Next, the area enclosed within each ROI was measured (Fig. 2B). The variation in ROI in the 12 images over the cardiac cycle is depicted in Fig. 2C.
Next, 12 maximum velocities were collected after scan, as demonstrated in Fig. 2B.
Because we found that the aqueductal area of the rst scan(which was de ned as A 1 ) was larger than the ninth(which was de ned as A 9 ) statistically signi cant, then the rst max velocity(V max1 )were chose to compare with the ninth V max9 .
Lastly, the SV was recalculated based on a dynamic ROI. The rst area ROI from each cardiac cycle was used to calculate the rst SV, which was de ned as SV 1 . Similarly, the ninth area ROI was used to calculate the ninth SV, de ned as SV 9 . Then compare SV 1 with SV 9 . The speci c process is shown in Statistical analysis SPSS ver. 19.0 software (SPSS, Inc., Chicago, IL, USA) and R ver. 3.6.0 software were used for the statistical analysis. All data were presented as the mean ± standard deviation (SD). To analyze the change in area and maximum velocity of the aqueduct during a cardiac cycle, the change in area and maximum velocity of the 12 observations for each patient was de ned as the difference between the area and maximum velocity at each timepoint and the initial ROI area. A t-test with a Bonferroni correction was then applied to compare the variation in ROI area and maximum velocity with time, separately. Applying Bonferroni correction, P = 0.05 was divided by the number of tests (11) to get the Bonferroni critical value, with P < 0.05/11 de ned as signi cant.
The student's t-test was used to compare A 1 vs A 9 , V max1 vs V max9 , SV 1 vs SV 9 . A P-value of < 0.05 was considered statistically signi cant.

Results
The nine volunteers included four men and ve women age reange 21-57 yrs, mean ± std 29.6 ± 9.1 yrs. All patients had no prior history of neurologic disease, no cerebrovascular disorders, and no medication use.
The aqueduct was scanned 12 times during a cardiac cycle, based on the R-R interval. The changes in area of the aqueduct throughout the cardiac cycle are displayed in Table 1. P value at each time point is shown in the last column of the table below, and 3 of them show a signi cant P value (P < 0.05). After Bonferroni correction was applied, the difference between the measures at the ninth time point and the rst time point is signi cant (P < 0.05/11). (Table 1, Fig. 4) The changes of maximum velocity in the aqueduct throughout the cardiac cycle are displayed in Table 2. P value at each time point is shown in the last column of the  The results of this study showed that V max1 was signi cantly lower than V max9 (-1.16 ± 1.89 vs -5.77 ± 1.89 P = 0.001) (Fig. 6). Compare with S1 and S9,the difference is signi cant(81.56 ± 20.17 vs 68.22 ± 16.02 P = 0.004) (Fig. 7). The results also showed that max velocity is inversely proportional to the area of aqueduct.

Discussion
The ROI size affects the mean velocity, ow rate, and SV differently. The ow in the center is higher than that at the periphery, so a small ROI placed in the center of the aqueduct will give higher readings of mean velocity while a larger ROI will underestimate it [24] . The max velocity representing the fastest ow in any pixel in the given ROI is not affected by the ROI size [25] . As a result, we chose maximum velocity for our analysis.
Karin Markenroth Bloch et al [18] reported that the speed curve conforms to the sinusoidal curve. Our results are consistent with his research. This study showed that S 1 is signi cantly larger than S 9 in a cardiac cycle, as a result, we chose V max1 and V max9 to compare, and we found that the maximum velocity may be inversely proportional to the area of the aqueduct. Dieter R. Enzmann et al [26] reported that the primary driving force behind intracranial and spinal canal CSF ow is expansion of the brain during vascular systole since arterial in ow and venous out ow are not equal throughout a cardiac cycle, and there is a short period of brain expansion during vascular systole. According to Dieter R. Enzmann's theory, we speculate that the brain tissue expands during the arterial systole and squeezes aqueduct, then aqueduct becomes smaller; at the same time the intracranial pressure increases and the ow rate of cerebrospinal uid becomes faster. In contrast, during arterial diastole, brain tissue will shrink, the force of the surrounding the aqueduct squeezed by brain tissue is reduced, the area is expanded, while the intracranial pressure is reduced and the ow rate of cerebrospinal uid is slowed.
SV is an important parameter often used in PC-MRI research, especially in aqueducts. SV is sometimes calculated differently because of the evaluation software chosen. For example, SV was averaged over the diastolic and systolic fluxes in the paper of Bradley et al [11] . Shanks J de ned SV as the volumetric mean of the caudal and cranial ow of CSF through the aqueduct [14] . In our study, SV of the aqueduct is equal to the net ow of cerebrospinal uid during a cardiac cycle (approximately forward flow volume minus backward flow volume), similarly to a prior study by Sartoretti et al [5] . Cerebrospinal uid ow within the aqueduct is best described as a to-and-fro motion with a very small net ow, and this normal variation is mainly related to the size and anatomy of CSF spaces, systolic and diastolic arterial blood pressure, jugular venous ow, and respiration [27] .
In most of the current post processing software, SV is equal to Mean flux (ml/s)×60/ heartbeat (1 RRinterval). = Mean velocity × Area of ROI. Theoretically, mean velocity becomes smaller as the Area of ROI becomes larger, so, mean flux remains unchanged. As a result, various size of manually delineated ROI appeared in many studies [28] . However, some research shows that there remains a considerable inaccuracy in the volume data effected by the placement of the ROI 29,30,31] , but these studies did not point out speci c reasons. The results of this study also show that volume data are affected by the placement of the ROI. Because we choose the area of the aqueduct at different times as a reference for ROI, maybe the reason of volume data affected by the placement of the ROI is related to the change of aqueductal area.
The de nition of SV 1 and SV 9 is based on the area of the aqueduct, while the latter is determined by the area of the ROI. Whether SV 1 or SV 9 is closer to the real ow requires further study. Anyway, the difference between SV 1 and SV 9 is signi cant indicating that volume data are affected by the placement of the ROI.
Therefore, we should consider how to de ne the ROI more accurately when detecting SV. In addition, whether there are other clinical signi cance remains to be further studied.
Because the ow rate of CSF in the aqueduct is very slow [22] . It is a challenge to quantitatively detect it using PC-MRI. Najafi et al reported that PC-MR was able to quantify low flow rates in vitro (0.1-5 ml/s) with a maximum underestimation of 5-10% [29] . The PC-MR imaging parameters chosen in this study are similar to theirs. Many parameters affect SV. One important parameter of the PC sequence is the VENC. Low VENC demands large gradients, resulting in long TE, which can increase signal loss due to intravoxel dephasing, especially in pathological situations with complex ow [18] .The value of the VENC should be chosen as close to the velocity of the fluid flow encountered in subjects [5] . A relatively high VENC (12 cm/s) was chosen in this study because the subjects presented with peak velocities of around 2-11 cm/s. This study has several limitations: rst, the number of volunteers is small, especially when comparing SV 1 and SV 9 , which may affect the results. Secondly, 2D PC-MRI was used in this study, and velocity was encoded in only one spatial direction, which resulted in in-plane velocity images or through-plane ow curves [22] . Because the motion of CSF through the ventricular system is a complex three-dimensional dynamic process, 3D analysis may be more accurate. Third, the effects of some related factors on breathing, sleep, and age have not been analyzed, which maybe have in uence on results [5,32,33] .

Conclusions
There is variation in the size of the cerebral aqueduct which is used to calculate stroke volume and other CSF ow parameters during the cardiac cycle. The maximum velocity may be inversely proportional to the area of the aqueduct. In order to establish reliable reference values for CSF ow parameters in future studies, a variable ROI, to account for cardiac cycle variation, should be considered and incorporated.

Declarations Acknowledgements
We thank Xiaobing Cheng and volunteers for assistance in the study.
Authors' contributions HZ: experiment design and conception, data collection, data analysis, manuscript writing; WD: data analysis, manuscript revision; XPL: data analysis, manuscript revision; YW: data analysis, manuscript revision; XYL: data analysis, manuscript revision; PMC: data analysis, manuscript revision; BDE: experiment design and conception, data collection, data analysis, manuscript revision. All authors read and approved the nal manuscript. All reasonable requests for data will be gladly granted by the corresponding author.
Ethics approval and consent to participate The study protocol was approved by the The First A liated Hospital of Henan University of Science and Technology Ethics Committee and informed consent was obtained from all participants.
[32]. Vinje V, Ringstad G, Lindstrøm EK, Valnes LM, Rognes ME, Eide PK, et al.   Change in aqueduct size during a cardiac cycle. A: ROI was drawn and it was enclosed the aqueduct with as little surrounding tissue as possible. B: The aqueduct Area and Max Velocity of each scan was automatic calculated. C: All of ROI were drawn with the same criterion.

Figure 3
Illustration of stroke volume measured with different ROI areas, xed over the course of the cardiac cycle.
A: All of ROI were the same according to area we chose. B: ROI was drawn around the center of the aqueduct. C: The areas of all scanned are the same. D: Parameters of absolute stroke volume (ml), forward ow volume (ml), backward ow volume (ml), regurgitation fraction, mean ux (ml / sec), stroke distance (cm), mean velocity (cm / sec) are show with one ROI area.
Page 16/18  Illustration of the changes in max velocity cm/s of the aqueduct throughout the cardiac cycle.

Figure 8
Illustration different results of SV1, and SV9, which uses a unique ROI for each phase of the cardiac cycle. SV1 was signi cantly lower than SV9. * P < 0.05; ns, not signi cant.