Abstract
Quantifying the effects of free breathing on cerebral venous flow is crucial for understanding cerebral circulation mechanisms and clinical applications. Unlike conventional cine phase-contrast MRI sequences (CINE-PC), real-time phase-contrast MRI sequences (RT-PC) can provide a continuous beat-to-beat flow signal that makes it possible to quantify the effect of breathing on cerebral venous flow. In this study, we examined 28 healthy human participants, comprising of 14 males and 14 females. Blood flows in the right/left internal jugular veins in the extracranial plane and the superior sagittal sinus (SSS) and straight sinus in the intercranial plane were quantified using CINE-PC and RT-PC. The first objective of this study was to determine the accuracy of RT-PC in quantifying cerebral venous flow, relative to CINE-PC. The second, and main objective, was to quantify the effect of free breathing on cerebral venous flow, using a time-domain multiparameter analysis method. Our results showed that RT-PC can accurately quantify cerebral venous flow with a 2 × 2 mm2 spatial resolution and 75 ms/image time resolution. The mean flow rate, amplitude, stroke volume, and cardiac period of cerebral veins were significantly higher from the mid-end phase of expiration to the mid-end phase of inspiration. Breathing affected the mean flow rates in the jugular veins more than those in the SSS and straight sinus. Furthermore, the effects of free breathing on the flow rate of the left and right jugular veins were not synchronous. These new findings provide a useful reference for better understanding the mechanisms of cerebral circulation.
Significance Statement
Currently, the effects of free breathing on cerebral venous flow are not well-understood. In this study, we first validated the accuracy of real-time phase-contrast sequences in quantifying cerebral venous blood flow. Subsequently, we used a novel analytical method to confirm and quantify the effects of free breathing on various parameters of cerebral venous flow across both intracranial and extracranial planes. We found that the influence of free breathing varies between these planes, and its effects on the cardiac period and blood flow are asynchronous. These novel findings provide a new perspective for understanding cerebral circulation mechanisms, particularly cerebrospinal fluid (CSF) circulation, and offer valuable insights for advancing clinical research and treatment strategies.
Introduction
Many neurological diseases affect the cerebral circulation (blood flow and oscillations of CSF) before they cause morphological abnormalities in the brain (Claassen et al., 2021; Drew, 2022). Hence, a better understanding of the physiopathological mechanisms of cerebral blood flow and CSF oscillations is essential for the early diagnosis and treatment of many neurological disorders.
Conventional cine phase-contrast MRI (CINE-PC) is the gold standard technique for quantifying cerebral neurofluid circulation and is widely used in clinical and basic research (Nayler et al., 1986; Ayadi et al., 2022). Along with the cardiac system, breathing can affect the cerebral circulatory system (Morgan et al., 2010; Skytioti et al., 2017). However, the characteristics of the cardiac cycle flow curve (CCFC) are averaged over all the cardiac cycles that have occurred during the acquisition (Pelc et al., 1991). The average CCFC determined by CINE-PC makes it difficult to quantify the effect of breathing on cerebral fluid circulation.
In recent years, real-time phase-contrast MRI (RT-PC) has been used to quantify cerebral circulation (Chen et al., 2015). Although RT-PC has not yet been applied in clinical practice, this technique appears to have great potential for quantifying the effects of breathing on cerebral fluid circulation and for clinical diagnosis (Dreha-Kulaczewski et al., 2015; Balédent et al., 2019; Lloyd et al., 2020; Baselli et al., 2022; Kollmeier et al., 2022). However, most of these studies focus primarily on quantifying the effects of specific breathing patterns on CSF. In contrast, the effects of free breathing on cerebral circulation, although more representative, is more challenging to observe (Laganà et al., 2022a,b). Furthermore, the considerable variability in the morphology of cerebral veins, especially in the extracranial cross-section of the jugular vein, has posed limitations on our current understanding of how free breathing affects cerebral venous blood flow.
Two primary methods can quantify the effects of free breathing (after referred to as breathing effects) on cerebral fluid circulation. Firstly, frequency-domain analysis quantifies the amplitude ratio or power ratio between the flow signal's heart rate components and its respiratory rate components (Skytioti et al., 2017; Baselli et al., 2022; Laganà et al., 2022b). However, this method can only quantify the impact of breathing on the mean flow rate. Secondly, time-domain analysis uses a belt sensor to record respiratory signal during MRI acquisition, segmenting the flow signal into independent CCFCs. Inspiratory and expiratory CCFCs are identified and compared to quantify the breathing effects on multiple variables of blood flow. It is important to note that the phase shift of breathing effects must be considered for accurate quantification as direct comparison of blood flow variables during expiration and inspiration might underestimate the intensity of breathing effects (Laganà et al., 2022b). For example, during free breathing, the peak cardiac period typically occurs between late expiration and early inspiration (Hirsch and Bishop, 1981; Galletly and Larsen, 1998). A prior research introduced a novel time-domain analysis approach to quantify the breathing effects on cerebral arterial blood flow parameters, considering phase shifts (Liu et al., 2022b).
The first objective of this study was to evaluate the accuracy of RT-PC in quantifying cerebral venous blood flow compared to the gold standard, CINE-PC. The second and main objective was to quantify the breathing effects on cerebral venous blood flow using RT-PC.
Material and Methods
Study conditions and participants
This study was approved by the local investigational review board (CPP Nord Ouest II, Amiens, France; reference: PI2019_843_0056) and was performed in accordance with the Declaration of Helsinki.
Twenty-eight human participants were included in the study (14 women and 14 men; mean ± standard deviation (SD) age: 26 ± 3.8; age range: 19–35). The MRI examination lasted about 30 min and was preceded by an interview, during which each patient's eligibility was assessed. All participants received information about the study's objectives and procedures, and gave their written consent to participation. The main exclusion criteria were a contraindication to MRI and a history of cerebrovascular or respiratory disease.
Image acquisition
A 3T MRI system (Philips Achieva; maximum gradient = 80 mT/m; slew rate = 120 mT m−1 ms−1) equipped with a 32-channel head coil was used to acquire images of participants placed in the supine position.
Each participant's intracranial and extracranial acquisition planes were localized by three-dimensional phase-contrast angiography [echo time (TE)/repetition time (TR) = 3 ms/5 ms; field of view (FOV) = 350 × 350 × 350 mm3; spatial resolution = 1.5 × 1.5 × 3 mm3; flip angle = 12°; number of acquisitions = 3], and both planes were then imaged using CINE-PC and RT-PC (Fig. 1). The extracranial plane contained the right internal jugular vein (RJ) and the left internal jugular vein (LJ), and the intracranial plane contained the superior sagittal sinus (SSS) and straight sinus (SS). The direction of blood flow to the heart was defined as positive and appears in black on the phase images.
Images acquisition (A) and Image processing (B) of CINE-PC and RT-PC. A, The intracranial plane (top) contained the straight sinus (SS) and the superior sagittal sinus (SSS), and the extracranial plane (bottom) contained the right internal jugular vein (RJ) and the left internal jugular vein (LJ). The amplitude images (left) and phase-contrast images (right) for the CINE-PC (left) and RT-PC (right) methods are shown. The green dashed line corresponds schematically to the venous trajectory. B, An example of segmentation of the RJ on CINE-PC images (in blue) and RT-PC images (in red). The flow rate signals from CINE-PC (B1: an averaged CCFC with 32 sampling points) and from RT-PC (B2: a continuous flow rate signal with 600 sampling points). The red points in the RT-PC signal represent the minimum value of each CCFC. (B3) The reconstructed mean ± SD CCFC (red), using all CCFCs from the RT-PC signal. (B4) A comparison of the differences in the three parameters between the CCFC from CINE-PC and the reconstructed CCFC from RT-PC.
Both CINE-PC and RT-PC were cartesian trajectory, with parallel acquisition technique [sensitivity encoding (SENSE) (Aja-Fernández et al., 2014)], FOV = 140 × 140 mm2, velocity encoding = 60 cm/s. The phase-contrast image of these two sequences was calculated by subtracting two velocity maps obtained with opposite bipolar gradients (i.e., an opposite-polarity flow-encoded pair). The CINE-PC in this study had a spatial resolution of 1 × 1 mm2, a thickness of 2 mm, a flip angle of 30°, an echo time of 6.6 ms, and a repetition time of 10.9 ms. The RT-PC had a spatial resolution of 2 × 2 mm2, a thickness of 4 mm, a flip angle of 10°, an echo time of 4.9 ms, a repetition time of 9.4 ms, and an EPI factor of 7. In addition, the SENSE factor was higher in the RT-PC (2.5) compared to the CINE-PC (1.5).
Firstly, we used a gradient-recalled echo CINE-PC sequence. A finger plethysmograph was used for cardiac gating. Each CINE-PC acquisition provided 32 phase-contrast images representing blood flow changes over an average cardiac cycle. Depending on the heart rate, the acquisition time ranged from 47 to 137 s.
Secondly, we used a multishot, gradient-recalled echo-planar imaging (EPI) RT-PC sequence. The EPI factor indicates the number of phase encodings filled in K-space for each TR. In the present study's conditions, the K-space of each velocity map had to be filled by four TRs; in other words, there were four segments (number of segments = the number of phase encodings divided by the SENSE factors and the EPI factors). Since each phase-contrast image, obtained by subtracting one velocity map from another, required 8 TRs, the time resolution of RT-PC was 75 ms (8 × TR). The acquisition time of RT-PC was influenced by the number of images. Here, we acquired 400 images for the first 16 participants (acquisition time = 30 s) and 600 for the last 12 participants (i.e., acquisition time = 45 s). To address the study's secondary objective, breathing signals during the acquisitions were measured (using a chest belt) for 21 participants. All participants breathed freely during the acquisitions.
Image processing
CINE-PC and RT-PC data were post-processed using Flow software, in three steps (Fall et al., 2019; Liu et al., 2022c).
Segmentation
The static region of interest (ROI) containing each vein was segmented automatically. The software also performed semi-automatic dynamic segmentation or manual segmentation, in order to adjust for vessel deformation or correct for movement by the participant.
Background field correction and aliasing correction
The background field was corrected by a fully automatic algorithm that defined the stationary tissue ROI around the target vessel. Aliasing artefacts caused by blood flow velocities greater than the velocity encoding value were corrected by the software's integrated de-aliasing algorithm (Liu, 2021).
Extraction of blood flow rate signals
The corrected blood flow rate signals for CINE-PC (an average CCFC with 32 sampling points) and RT-PC (a continuous blood flow rate signal containing 400 or 600 sampling points and multiple CCFCs) were then extracted into the software's database for subsequent processing and analysis.
RT-PC vs CINE-PC
To compare the two sequences, the continuous flow rate signal from RT-PC was reconstructed to obtain a CCFC similar to that of CINE-PC. The software's algorithm first located the minimum values of each CCFC in the continuous blood flow signal and then divided the signal into multiple independent CCFCs, using the minimum values as segmentation points (Fig. 1B2). Next, the number of sampling points was increased to 32 by interpolating each CCFC. Lastly, all CCFCs were reconstructed into a single CCFC with 32 sampling points (Fig. 1B3).
In the present study, CINE-PC was used as the “gold standard” against which the accuracy of RT-PC was measured. Hence, we compared the CCFCs derived respectively from RT-PC and CINE-PC with regard to the area segmentation, the mean flow rate (average of 32 sampling points) and the pulsatility index [the amplitude of the CCFC divided by the mean flow rate (Fig. 1B4)].
Quantification of the breathing effects on cerebral venous flow
The multiparameter time-domain analysis of the breathing effects on cerebral venous flow comprised five steps:
Definition of inspiratory and expiratory phases.
The inspiratory and expiratory phases were defined as the rising and falling parts of the breathing signal, respectively. The inspiratory phase corresponds to the red area in Figure 2A, within which the respiratory signal was evaluated between the minimum value and the maximum value (the red arrow in Fig. 2).
Reconstruction of the CCFCEx (during expiration) and CCFCIn (during inspiration).
Flow chart for quantifying the breathing effects (DiffEx-In and Shift) on cerebral venous flow. A, Definition of the inspiratory phase (IN, red) and the expiratory phase (EX, blue), using the breathing signal. B, Reconstruction of CCFCEx (the blue curve) and CCFCIn (the red curve) from the respective expiratory CCFC (blue points) and inspiratory CCFC (red points) extracted from the RT-PC flow signal. C, Calculation of DiffEx-In for each variable evaluated, with zero phase shift (Shift = 0 s). D, The respiratory signal shift from −3 s to +3 s in increments of 0.1 s. The preceding steps are repeated in order to obtain the DiffEx-In (p, shift) curves for the four variables evaluated. E, Extraction of the maximum value of each DiffEx-In (p, shift) curve, in order to estimate the intensity DiffEx-In and the Shift (%) of the breathing effects on each variable.
Based on the mid-point of each CCFC, the software determined whether the CCFC was measured during the inspiratory or expiratory phase (Fig. 2A). Next, using the same reconstruction approach, all CCFCs obtained during the expiratory phase were reconstructed into an average CCFCEx. The same was done for CCFCIn (Fig. 2B).
Quantification of the percentage difference between CCFCEx and CCFCIn, for each variable.
Based on the CCFCEx and CCFCIn for each cerebral vein, four variables p (mean flow rate, amplitude, stroke volume, and cardiac period) were calculated (Fig. 2B):
The mean flow rate refers to the average value derived from 32 sampling points during the reconstructed CCFC. The amplitude is defined as the difference between the maximum and minimum values observed in the reconstructed CCFC. The stroke volume is calculated as the integral of the CCFC, which represents the volume of venous blood flowing towards the heart during each heartbeat. The cardiac period is determined by the time interval that spans from the first to the last sampling point of the CCFC. The percentage difference between CCFCEx and CCFCIn for each variable p (DiffEx-In (p)) was calculated using the equation below (Fig. 2C).
Generation of the DiffEx-In (p, shift) curve
To take account of a potential phase shift in the breathing effects, we applied a shift variable to move the inspiratory and expiratory phases. The first three steps in the multiparameter time-domain analysis were then repeated, in order to calculate the value of DiffEx-In (p, shift). Considering that the period of free breathing was usually less than 6 s, the range of shift values was set from −3 s to +3 s in increments of 0.1 s. The DiffEx-In (p, shift) curve was obtained after 60 iterations for the four variables evaluated in each vein or sinus.
If free breathing had an effect, the period of the DiffEx-In (p, shift) curve should be similar to that of the breathing signal (Fig. 2D).
Estimation of the intensity DiffEx-In, and the percentage shift Shift (%) of breathing effects on cerebral venous variables.
The maximum value of the DiffEx-In (p, shift) curve (denoted as DiffEx-In) indicates the intensity of breathing effects on the venous variables. To enable comparisons of participants with different breathing periods, the percentage of the corresponding shift (in s) with respect to the mean breathing period was recorded as shift (%) (Fig. 2E).
After quantifying the breathing effects, the “ΔShift” variable is used to quantify the phase difference between the breathing effects on RJ and the breathing effects on the LJ. ΔShift ranges from 0% to 50% and corresponds to the absolute difference between the Shift of RJ and Shift of LJ. A smaller ΔShift indicates that the breathing effects on RJ and on LJ are more synchronized. Furthermore, Δarea was also calculated as the absolute percentage area difference between RJ and LJ. The correlation between Δarea and ΔShift was evaluated.
Statistical analysis
The statistical analyses were performed with R software (R Core Team, 2020). Descriptive statistics were reported as the mean ± SD. The Shapiro–Wilk test was used to determine whether or not the data were normally distributed. Student's t-test or Wilcoxon's test (depending on the normality of the distribution) was used to detect differences between pairs of groups. Intergroup correlations were assessed with Pearson's test (for normal distributions) or Spearman's test. A Bland–Altman plot (Martin Bland and Altman, 1986) was used to quantify the degree of agreement between RT-PC and CINE-PC. All statistical tests were two-sided, and the threshold for significance was set to p < 0.05.
Results
RT-PC vs CINE-PC
Full datasets from two of the 28 participants were excluded due to uncorrectable motion and aliasing artifacts. For two of the remaining 26 participants, two extracranial plane datasets were then excluded because of uncorrectable aliasing artifacts. Therefore, 26 intracranial plane datasets and 24 extracranial plane datasets were assessed for the primary objective.
The values of the cardiac period, area, mean flow rate, and pulsatility index for the four veins and sinuses, as quantified by CINE-PC and RT-PC sequences are shown in Table 1 and Figure 3. Of the 24 extracranial planes, one showed the LJ only and five showed the RJ only.
Mirrored density plot of quantitative vessel results for CINE-PC (blue) and for RT-PC (red). The three lines in the density diagram represent the interquartile range and the median. The significance of difference in the means for CINE-PC vs RT-PC or for Extra vs Intra were determined in a paired Student's t-test (*p < 0.05; **p < 0.01). RJ, right internal jugular vein; LJ, left internal jugular vein; SS, straight sinus; SSS, superior sagittal sinus; Extra, extracranial plane; Intra, intracranial plane.
Comparison of the four veins and sinuses, as quantified by CINE-PC and RT-PC sequences
There is a strong correlation between the results obtained by CINE-PC and RT-PC for the three variables and the four vessels. The correlation coefficient r ranged from 0.55 to 0.98 (p < 0.01, Pearson's test, in all cases). The segmented areas of LJ, SS, and SSS measured by RT-PC were significantly larger than those measured by CINE-PC (by 8.4%, 20%, and 4.7%, respectively). The mean flow rates in the SS and SSS measured by RT-PC were significantly larger than those measured by CINE-PC (by 7.5% and 3.5%, respectively). For pulsatility index, only the SSS gave a significantly higher value in RT-PC than in CINE-PC (by 10%). In both CINE-PC and RT-PC, the mean flow rate was significantly higher in the extracranial plane than in the intracranial plane (20% for CINE-PC and 13% for RT-PC). For all three variables, the degree of inter-individual variability was lower for the SS and SSS than for the RJ and LJ (Fig. 3).
When considering all cerebral vessels (the left-most column of the Bland–Altman plots in Fig. 4), the area segmentation gave the largest mean percentage difference (10.6%), the pulsatility index presented the largest limits of agreement (LOA) (−54.8% to 40.4%), and the mean flow presented the smallest mean percentage difference (2.2%) and LOA (−18.3% to 22.6%). For all three variables, the degree of inter-individual variability was lower for the SS and SSS than for the RJ and LJ (Figs. 3, 4).
The Bland–Altman plots for RT-PC vs CINE-PC and the three variables evaluated (from top to bottom: the segmented area, the flow rate, and the pulsatility index). The x-axis corresponds to the mean value of RT-PC and CINE-PC, and the y-axis corresponds to the percentage difference between RT-PC and CINE-PC. The red dotted line indicates the mean percentage difference, and the top and bottom dashed lines indicate the 95% LOA (calculated as the mean ± 1.96 SD).
Quantification of the effects of breathing
Datasets from three participants were excluded because of erratic breathing during the acquisition. Hence, datasets from 18 participants were used to evaluate the study's secondary objective. The mean cardiac period was 0.89 s ± 0.19 s (range: 0.57–1.5 s), and the mean breathing period was 3.75 s ± 0.67 s (range: 2.82–5.43 s).
As shows in Table 2, the DiffEx-In (%) was significantly lower when the phase shift was disregarded. Furthermore, upon considering the phase shift, Table 2 shows that firstly, with regard to amplitude, DiffEx-In was the largest and Shift presented the highest variability. Secondly, the breathing effects (DiffEx-In and Shift) on mean flow, stroke volume and cardiac period for SS and SSS were correlated (p < 0.01 in Pearson's or Spearman's test). Comparing the RJ and LJ, the breathing effects were only correlated for the cardiac period (p < 0.01, Pearson's test). Thirdly, DiffEx-In on the mean flow were significantly higher in the RJ and LJ (6.9% and 6.3%, respectively) than in the SS and SSS (2.7% and 2.8%, respectively). Fourthly, the ΔShift on the cardiac period was smaller than the ΔShift on the mean flow. Fifthly, ΔShift for the mean flow in the RJ–LJ was larger than the value in SS–SSS (25% and 11%, respectively). The distribution of DiffEx-In and Shift values for each variable can be found in Figure 5.
Distribution of the values of DiffEx-In (top) and Shift (bottom) for each variable (by color) and each cerebral vein or sinus. The line plot shows the correlations for the RJ vs LJ and the SS vs SSS. *p < 0.05, **p < 0.01, in a paired Student's t-test.
Quantification of the breathing effects (intensity DiffEx-In with and without Shift) on four variables in four cerebral veins or sinuses
The relationship between Δshift on mean flow and Δarea for the RJ and LJ is shown in Figure 6A. As the Δarea increases, the Δshift also increases (p < 0.01, Pearson's test), the details of two specific data points are shown in Figure 6B and C. Figure 6B1 and C1 illustrates the filtered flow rate signals (breathing bands) for the RJ and LJ obtained in a frequency-domain analysis. Figure 6B2 and C2 presents the DiffEx-In (mean flow, shift) curves obtained with the time-domain analysis method. It is noteworthy that the respective Δshift values obtained with the frequency and time-domain analyses are similar, which emphasizes the reliability of our method. A schematic diagram of the filtering process is given in Figure 6B.
The correlation between Δshift (of breathing effects on mean flow for RJ–LJ) and Δarea (for RJ–LJ). In (A), the x-axis indicates the Δarea (%) and the y-axis indicates the Δshift (%). B,C, Show the details of the two datapoints selected in A. B1 and C1 are bandpass (breathing band)-filtered flow rate signals from the RJ and LJ. B2 and C2 show the DiffEx-In (mean flow, shift) curves for the RJ and LJ.
The distribution of breathing effects (Shift and DiffEx-In) values for the mean flow and cardiac period of the SSS is shown in Figure 7. As can be seen in Figure 7A (mean flow), Shift is more variable and is significantly and negatively correlated with DiffEx-In (p < 0.01, Pearson's test). Figure 7B and C shows that Shift and DiffEx-In on mean flow are correlated with the breathing period. As the breathing period increases, DiffEx-In also increases (p = 0.02, Spearman's test), while Shift tends to decrease (p < 0.01, Spearman's test).
The distribution of the breathing effects values for the mean flow and cardiac period of the SSS (A), and the influence of the breathing period on the breathing effects on mean flow in the SSS (B,C). The r and p in (A) correspond to Pearson's coefficient and the p-value for the correlation between the Shift and DiffEx-In (mean flow). The r and p in (B,C) correspond to Spearman's coefficient and the p-value for the correlation between the breathing period on one hand and DiffEx-In (in B) and Shift (in C) on the other.
Discussion
Firstly, we confirmed the feasibility and reliability of RT-PC for quantifying cerebral venous blood flow. Secondly, we used a multiparameter time-domain analysis (with the addition of the Shift variable) to quantify the breathing effects on cerebral venous flow.
RT-PC vs CINE-PC
The geometric distortion artifacts caused by the long readout time with EPI sequence have always constituted a challenge. In the present study, we used a multishot EPI RT-PC sequence; the corresponding readout time fell to the level seen for CINE-PC (Heidemann et al., 2010) and helped to avoid geometric distortion artifacts (Fig. 1).
The spatial resolution corresponds to the acquisition pixel size: 2 × 2 mm2 for RT-PC in this study, and 1 × 1 mm2 for CINE-PC. The pixel size affects (i) the velocity-to-noise ratio (which is lower in RT-PC than in CINE-PC because several echoes (EPI factor = 7) are acquired in one TR; and (ii) the time resolution in RT-PC. In the case of CINE-PC, the pixel size only affects the acquisition time; thanks to cardiac gating, the sequence can fill a larger acquisition matrix by adding more cardiac cycles while keeping its pseudo-time resolution constant (32 images for an average CCFC). Therefore, we used a larger pixel size for RT-PC to increase the velocity-to-noise ratio and the time resolution.
Nevertheless, it is important to note that increasing the pixel size too much can cause a partial volume effect; this resulted in a slightly larger area segmentation for RT-PC than for CINE-PC (by 10.6%). Especially for the smaller SS, the percentage difference in area segmentation between RT-PC and CINE-PC reached 20.3% (Fig. 4). This overestimation of area segmentation can also lead to overestimation of flow rates. Typically, the accuracy of flow rate quantification is more of a concern with phase-contrast sequences. Some researchers (Tang et al., 1995; Jiang et al., 2015) have shown that the ROI of CINE-PC needs to contain at least 12–16 pixels in order to maintain the flow rate error within 10%; the pixel size (1 mm2) in CINE-PC is sufficient for the present study. There are few literature data on the required number of pixels in an RT-PC ROI (Liu et al., 2022a). Although we observed a small partial volume effect in the SS, it had little impact on quantification of the breathing effects on cerebral venous flow (7.5% overestimation of flow rate).
CINE-PC does not depend on the heart rate and can always represent an average CCFC with 32 sampling points. In contrast, the number of CCFC sampling points in RT-PC is affected by cardiac period; a longer cardiac period can increase the accuracy of pulsatility index quantification by RT-PC. Differences in the heart rates during the RT-PC and CINE-PC acquisitions and inter-individual differences in the cardiac period might explain (at least in part) why the pulsatility index (Fig. 4) had a broader LOA in the Bland–Altman plot.
Quantification of the breathing effects on cerebral venous flow
Our present results confirmed that the phase shift (Shift) should be calculated before quantifying the intensity (DiffEx-In) of breathing effects. Compared with frequency-domain analysis, a multiparameter time-domain analysis avoids (to some extent) the impact of inhomogeneous breathing and arrhythmia and quantifies the breathing effects on many variables.
The effect of breathing on the cardiac period (also referred to as respiratory sinus arrhythmia) has been studied clinically for many years (Angelone and Coulter, 1964). Compared with amplitude and mean flow, Shift for the cardiac period was more homogenous (Table 2 and Fig. 5) and was located essentially in one-third of the breathing period (i.e., Shift = 33%). These results are consistent with other studies (Hirsch and Bishop, 1981; Galletly and Larsen, 1998; Yasuma and Hayano, 2004) in which the R-R interval on an ECG was longer at the end of expiration and shorter during inspiration.
The DiffEx-In in mean flow rates for extracranial jugular veins were more pronounced compared to those of intracranial sinuses (Table 2). This greater effect on the jugular veins could be attributed to their closer proximity to the heart and greater susceptibility to deformation, making them more responsive to the impact of breathing. In addition, cerebrovascular autoregulation may also contribute to stabilizing intracranial pressure, potentially resulting in relatively lower DiffEx-In in the sinuses. Conversely, certain veins not affected by cerebrovascular autoregulation (e.g., subcutaneous veins, facial veins) converge towards the jugular veins, thereby resulting in an increase of the DiffEx-In in the jugular veins. This hypothesis could be further verified in future studies by integrating the analysis of both arterial and venous cerebral blood flow.
Another interesting finding is that the Shift of the RJ was not correlated with that of the LJ (Table 2 and Fig. 5). However, a significant ΔShift between the RJ and LJ was shown (ΔShift of RJ–LJ = 25% vs ΔShift of SS–SSS = 11%). A larger ΔShift will result in a lower intensity (DiffEx-In) of breathing effects on the extracranial plane, therefore, RJ and LJ should be considered together when quantifying the breathing effects on the extracranial plane.
Moreover, the Δarea (of the RJ and LJ) was significantly correlated (p < 0.01) with ΔShift (of the RJ and LJ) (Fig. 6). Might morphological variability in the veins in the extracranial plane lead to a physiological adaptation in which (depending on the conditions) the venous arborescence adjusts in order to perform its function optimally? It is known that vascular area is correlated with vascular resistance (Fall et al., 2017). It would be interesting to look at whether this ΔShift is also correlated with the difference in vascular resistance between the RJ and LJ; this finding might have promising applications in clinical diagnosis.
By analyzing the breathing effects on the mean flow rate in the SSS, we observed that Shift was more variable than the cardiac period (coefficient of variation: 58.7% vs 25.6%, respectively). Furthermore, Shift was significantly and negatively correlated with DiffEx-In (Fig. 7A). This result is consistent with our previous work on the quantification of breathing effects on cerebral arteries (Liu et al., 2022b). Further analysis showed that the breathing period correlated with both DiffEx-In and Shift of breathing effects on mean flow in the SSS. As shown in Figure 7B and C, DiffEx-In increases and Shift decreases as with breathing period increases. One can reasonably hypothesize that due to the change in CO2 density in the blood flow caused by breathing, cerebral regulation changes the vascular resistance, adjusts the caliber of small blood vessels in the brain and thus regulating cerebral blood flow (Claassen et al., 2021). As the breathing period increased, the Shift on mean flow tended to decrease; in contrast, the Shift on the cardiac period remained quite stable, at 33% (Table 2). A longer breathing period would be associated with a longer period of cerebral autoregulation and thus a higher DiffEx-In. This finding need to be confirmed in further research but might become a useful variable for the clinical diagnosis of the brain's regulatory capacity.
Stroke volume is an important variable in the quantification of the cerebral blood volume and intracranial pressure. As shown in Table 2 and Figure 5, the breathing effects (DiffEx-In and Shift on stroke volume) on the RJ was not correlated with the effect on the LJ—mainly because the respective breathing effects on mean flow were also not correlated. In contrast, the breathing effects (DiffEx-In and Shift on stroke volume) in the SS and the SSS were well correlated; this confirmed that the breathing effects on the intracranial plane can be quantified by studying the SSS alone.
While this study yielded novel findings, there were limitations to be acknowledged. Abdominal breathing and chest breathing reportedly influence the breathing effects on cerebral CSF circulation (Aktas et al., 2019); however, the researchers did not specify the free breathing mode (abdominal or chest breathing). Due to equipment limitations, it was only possible to quantify the effect of free breathing on cerebral venous blood flow in the supine position, and the hemodynamic characteristics of cerebral venous blood flow may vary by different postures (Alperin et al., 2005; Laganà et al., 2017).
In future studies, an integrated analysis of arterial, venous, and CSF hemodynamics could be performed for a more comprehensive understanding of the influence of breathing on cerebral circulation. Additionally, using a higher sampling frequency finger plethysmograph signal to define the cardiac cycle, and interpolating blood flow signals, could enhance the sensitivity in quantifying variations within the cardiac cycle. As a result, this approach could improve the robustness and accuracy of CCFC segmentation.
In conclusion, our results demonstrated the accurate quantification of cerebral venous blood flow using RT-PC with a 2 × 2 mm² spatial resolution and 75 ms/image time resolution. We successfully quantified the intensity (DiffEx-In) and Shift of the breathing effects on four blood flow variables in two cerebral veins in the extracranial plane and two sinuses in the intracranial plane. These new findings are likely to constitute a valuable reference for further studies of the effects of breathing on the cerebral circulation.
Compliance with ethical standards
All procedures involving human participants were performed in accordance with the ethical standards of the institutional and/or national review board and with the 1964 Declaration of Helsinki and its amendments or comparable ethical standards. Informed consent was obtained from all individual participants involved in the study.
Footnotes
Acknowledgments: This research was funded by the French National Research Agency (reference: Hanuman ANR-18-CE45–0014 and EquipEX FIGURES 10-EQPX-0001) and INTERREG France (Channel) England Programme (the REVERT project). The authors thank David Chechin from Phillips Industrie for scientific support and Héléna Freulet, Garance Arbeaumont-Trocmé, Vilhem Marion and Julien Van Gysel (MRI research technicians) for assistance with the acquisition of high-quality images.
The authors declare no competing financial interests.
- Correspondence should be addressed to Pan Liu at liu.pan{at}chu-amiens.fr or Olivier Baledent at olivier.baledent{at}chu-amiens.fr.
