Using a prototype RT-PC (22) sequence with a temporal resolution of 58.5 ms for the blood and 94 ms for the CSF, we were able to estimate the blood and the CSF flow rates while the subjects breathed freely or with a constant rate, with a normal or forced respiration in the latter case. By computing the power spectrum of ICAs, IJVs and CSF flow rates in the three breathing conditions, we could study the respiratory and cardiac influences on the flows.
We previously presented this protocol in (24), but in the current study, acquisition during free breathing was additionally assessed, as it is the normal breathing condition during conventional PC-MRI. We also introduced an analysis of the normalized power in various frequency ranges, representing the respiratory and cardiac contributions to the flow rate variance. We also increased the sample size to 30 healthy subjects.
This setup allowed us to demonstrate that: 1) the LFP and the HFP of the spectrum and their harmonics corresponded to the respiratory and cardiac frequencies and their harmonics; 2) paced, and forced breathing on a greater extent, caused a blood mean flow rate decrement and an increment of respiratory modulation on all the fluids when compared to free breathing; 3) the fluid volumes displaced during inspiration vs expiration were not statistically different.
Flow rate drivers
The few previous works that estimated the venous and CSF flow rates using MRI in RT and computed their power spectrum described a LFP and a HFP in the spectra (12, 21, 22). The frequency of the latter matched to the typical cardiac frequency, while the former corresponded to the respiratory frequency asked to the subject during the experiment. We could demonstrate that the LFP and the HFP corresponded to the respiratory and cardiac frequencies measured during the RT-PC MRI acquisition with a thoracic band and a pulse oximeter respectively. The cardiac influence on blood and CSF flows is well known: periodic and typical oscillations have been shown and studied using the cardiac-gated PC-MRI that is routinely used (26). Respiration is the other main driver of blood and CSF flows, due to the pressure changes provided by the thoracic pump (4, 12, 22, 23, 27, 28), as shown by previous studies using RT-PC MRI (12, 22, 23, 28) and ultrasound (4, 27). The advantage of ultrasound over MRI in this kind of studies is that it is intrinsically an RT technique. Conversely, specific sequences have to be customised to perform RT-PC MRI acquisitions, being the commercially available PC sequence cardiac-gated. However, one of the advantages of MRI compared to ultrasound is that it allows the investigation of CSF besides blood flow rate. CSF respiratory oscillations have been enhanced using MRI in some previous works (11, 12, 20, 22, 23, 28). Only few recent MRI studies showed the respiratory influence on venous flow (11, 12, 29) too, and its changes with forced breathing. We decided to investigate the respiratory and cardiac influences on CSF, venous and arterial flow rates. To the best of our knowledge, only (11) studied the cardiac and respiratory influences also for the arterial flow, besides the venous and CSF flows. However, in that study an EPI sequence was used, rather than a PC-MRI sequence so the flow could not be quantified. Moreover, in our current study, a temporal resolution that is higher compared to that used in literature was adopted (11, 12, 20, 22, 23, 28).
Given our temporal resolution and image quality, our signals were rich of information, as it can be observed in Supplementary Fig. 1. This allowed us to verify also the presence of the second respiratory harmonic and the 2nd and 3rd cardiac harmonics. As shown for the fundamental harmonics (i.e. LFP and HFP), we verified that these other peaks were the respiratory and cardiac harmonics, corresponding to 2*BR, 2*HR, and 3*HR. The differences we found between the 2nd harmonic of the LFP and the 2*BR for the ICA and CSF in the F condition have to be probably ascribed to low peaks amplitude in F breathing, which may have produced peak misdetection (i.e. a peak of a similar height but of slightly different frequency could have been detected near to the real 2nd harmonic LFP). The differences detected between the 3rd HFP harmonic of the ICA spectrum in the PN and 3*HR could be ascribed to the same reason.
Breathing type influence on mean flow rates
As recently shown by Kollmeir et al. (12) in 16 healthy subjects, we confirmed the venous flow rate decreases with forced breathing. We also showed that paced breathing compared to free breathing, and forced paced breathing compared to a normal paced respiration, provoked arterial flow rate decrement. The arterial flow rate decrement was concurrent to ICAs cross-sectional area decrement, so it was a consequence of vasoconstriction. Vasoconstriction probably occurred because an increase of oxygen saturation due to respiration strength increment induces a vasocompensatory effect on arteries and reduces arterial flow rate. This physiologic flow autoregulation helps maintaining constant brain oxygen. As a consequence of arterial flow rate decrement with paced breathing, we observed that also concurrent IJVs flow rate decrease, in absence of cross sectional area decrement.
Conversely to Kollmeir et al. (12), we did not obtain that the average CSF changes with breathing type. Indeed, the average CSF was near to zero, the average arterial flow decrement was well mirrored by venous decrement, but not by a CSF average change.
Since we expected the blood flow to decrease with forced respiration as in (12) and in our preliminary work (24), and since we wanted to compare the flow rates in three breathing conditions, we designed our study protocol so that each RT-PC acquisition with paced breathing started three breathings after we asked to the subject to breathe with our pacing. Our preliminary tests (Supplementary Fig. 4) revealed that the mean flow rate decreased in a few seconds passing from free to paced breathing, and that after three complete breathing cycles it reached a new regime. This can be slightly observed even when looking at the IJV flow rate over time represented in Fig. 3 of (12): the venous flow rate progressively decreased from 40s, when switching from free breathing to forced breathing, and it got lower and constant from 60 to 80 s (during forced paced breathing). In our current study transitory cycles were excluded from the analysis, so we could compare three “pure” breathing conditions. We acknowledge that our setup (excluding transitory cycles) does not allow investigating the change rate, while the setup of (12) would potentially allow that. If future studies examine the change rate in physiological or pathological cases, a continuous acquisition is recommended, for including different types of breathing in subsequence.
Breathing type influence on flow rate modulations
The frequency analysis showed that the respiratory component increased with pace breathing (both PN and PD) when compared to F breathing. Since the flow rate variance was stable during the different breathing types (results not shown), and since it is mainly dependent on the sum of the three cardiac components and the respiratory components, the sum of the powers of the cardiac peaks decreased with paced breathing when compared to free breathing, with an opposite and complementary trend compared to the respiratory power. This means that the flow rate oscillations at the respiratory frequency, i.e. the breathing modulations, become higher, while those at the cardiac frequency become lower in paced breathing with respect to free breathing. However, the cardiac contribution remained the prevalent modulation driver in all the types of breathing patterns, being the normalized cardiac power consistently higher than the respiratory one. The powers in the respiratory and cardiac ranges normalized by the signal variance are easy to interpret: they provide respiratory and cardiac flow rate modulations respectively with an index from 0 to 1. Interestingly, from Table 2 we can observe that the respiration accounts up to 50% of the flow rate variance for the IJVs in F and PN breathing, and had a maximum of 74% during PD breathing. In the latter breathing mode, the normalized power of respiration had a median value of 31%, similarly to the normalized cardiac power (30%). The respiratory normalized power during the PD breathing had a maximum of more than 40% for ICAs and CSF.
Deepening our frequency analysis with a specific evaluation of the three HFP harmonics, we observed (Supplementary Table 4) that the cardiac contribution decrement was due to the 1st and 2nd harmonic power decrement for IJVs and CSF; conversely it was due to a decrement of only the 1st harmonic for ICAs. Indeed, the normalized power in the first HR harmonic band in the PD breathing was significantly different compared to the other breathing modes for all the structures. Conversely, the normalized power in the second HR harmonic band did not change with breathing mode for ICAs, but was significantly lower in the PD compared to the F breathing for IJVs and CSF. This means that ICAs flow rate changed its shape with PD breathing compared to the other two breathing modes, while the oscillations at the HR and 2*HR both decreased for IJVs and CSF flow rates, not changing their main shape in the PD compared to the F breathing. This analysis also showed that the PD respiration increased the respiratory modulation so much that the normalized power of respiration was similar to the normalized power in the 1st (the highest) and to the 2nd (the second highest) cardiac harmonic for IJVs and CSF respectively. For ICAs, the normalized power of respiration in the PD breathing increased just until the normalized power of the 3rd cardiac harmonic (the smallest one) in the PD respiration. A similar result is shown by the R/C index: being it the ratio between the respiratory and the 1st cardiac powers, it was always below 1 for ICA and CSF, but it increased over 1 for IJV with paced breathing. Similarly, Kollmeir et al. (12) showed R/C below 1 for CSF during normal and forced breathing and for IJVs for normal breathing, but it exceeded 1 for some subjects during forced breathing.
The high respiratory modulation of the venous flow rate during the PD breathing confirmed the results of previous works (12, 23, 29), since the effect of the thoracic pump increases with forced respiration. Being the venous and CSF flows highly dependent (12, 26, 30), the respiratory modulation had a great effect also for the CSF flow rate, and increased with forced breathing when compared to other breathing conditions. We have to underline that with our experiment we demonstrated that the respiration type influences also the arterial flow.
Two peaks at the edges of the main HFP were also observed in paced breathing, and were particularly evident for the PD breathing. These are spurious harmonics, due to the interference of the two cyclic flow rates drivers, i.e. the respiration and cardiac beats.
Breathing phase influence on flow volume
We expected that during inspiration the venous flow volume was greater compared to expiration. We also expected that the CSF flow volume was directed cranially during inspiration, as suggested by some previous works (12, 23)}. However, this hypothesis was confirmed for IJVs only, during the paced breathing. Although the hypothesized result has been qualitatively reported in Kollmeier’s recent paper (Fig. 2) (12), in Akatas (23) (Fig. 3), and in Dreha-Kulaczewski (31) (Fig. 2), associated detailed statistics were not provided. This may justify the discrepancy between our results and the ones previously reported by Kollmeier (12), Akatas (23) and Dreha-Kulaczewski (31). In a previous study using ultrasound (4), we have reported that blood velocity in some of the examined major neck veins decremented during inspiration and incremented during expiration. Although respiration modulated the blood velocity in veins, different kind of phase synchronizations between flow rate and respiration were observed in real-time with ultrasounds. In the current study, we observed the same kind of result in IJVs, but also in CSF and ICAs. Specifically, only in some subjects the flow rate was synchronized with respiration so that the inspiration corresponded to increment in IJVs and ICAs and cranial CSF. For other subjects a time lag was observed, as the flow rate was not synchronously influenced by a particular respiratory phase. This probably prevented obtaining significantly different volumes between inspiration and expiration; indeed, only a trend of different volumes between inspiration and expiration was found for IJVs in PN breathing, for which most of the subjects had the same kind of alignment between respiration and flow rate.
Strengths and weaknesses
We have confirmed the recent results of Kollmeier et al.,(12) in a larger group of subjects: when comparing forced breathing to free breathing, the IJVs flow rate decreases, the respiratory modulation on cervical IJVs and CSF flow rates increases and the cardiac modulation decreases. Differently from the literature about similar measures using MRI, we focused also on the cervical arterial flow, showing that during forced breathing vasoconstriction occurs and that the ICAs flow rate decreases. This may be the probable driver of venous decrement for autoregulation.
However, our study has some limitations. As a general limitation of MRI RT-PC studies, the acquisitions are possible only in the supine position. However, that position deserves investigation, since the venous flow redistributes from IJVs to vertebral veins in the sitting position. Second, we separately acquired blood and CSF flows, so we could not investigate their temporal coupling during the various beats and respiratory cycles. However, acquiring both blood and CSF flows with a unique sequence with two subsequent VENC would have decreased the temporal resolution. Therefore, this experimental choice allowed reaching a higher temporal resolution than the one used in previous studies, for both the blood and CSF flow rates. Third, we did not measure the CO2 partial pressure changes during the various types of breathing, that would have helped the interpretation of the average flow rate change.
Another peculiarity of our study design is the inclusion of different kinds of breathing: we decided to ask the subjects to breathe not only in free and forced conditions, but also in paced condition with normal breathing strength. This allowed to confirm that forced breathing increases the respiratory modulations to flow rates, but also to evidence that the respiratory contribution can be enhanced with paced breathing.
We reported how breathing affects the arterial, venous and CSF average flow rates and their modulations in normal volunteers. Differently from previous studies, we described the power spectra normal values detailing different cardiac contributions: from the 1st to the 3rd HR harmonics. We computed the respiratory power normalized by the power in the 1st HR harmonic to compare our results with the same index used in literature, however we also introduced the normalized power in the various frequency range. This index between 0 and 1 represents the contribution of oscillations at a particular frequency range.
Considering that altered blood flow, and CSF net volume or peak velocity were described in many pathological conditions (19, 32, 33), the application of the current protocol into clinical studies might allow to investigate if there are differences in terms of how the flows in/out the brain are modulated by cardiac beating and by breathing patterns compared to the normal values. The current setup could also allow to study if breathing manoeuvre exacerbate CSF flow alterations, and if respiratory exercises allow to normalize blood and CSF flows. Indeed, forced breathing at a constant rate was proposed (34) as a clinical non-invasive manoeuvre for testing cerebral autoregulation, since it produced sinusoidal blood pressure oscillations at the respiratory frequency, then transmitted to the cerebral blood flow volume with different lags depending on the autoregulation response. There are further numerous possible clinical applications of this study, such as investigating the impact of respiration type for drug delivery through CSF (35, 36). Finally, this set-up could be used to study if blood and CSF circulations are altered in pathological respiration patterns, such as in patients with chronic obstructive pulmonary disease(37), or in obstructive sleep apnea (38).