| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1Department of Cardiovascular Sciences, Faculty of Medicine, University of Leicester, and 2Department of Medical Physics, University Hospitals of Leicester National Health Service Trust, Leicester, United Kingdom
Submitted 4 February 2005 ; accepted in final form 26 April 2005
 |
ABSTRACT |
|---|
 TOP ABSTRACT METHODSRESULTSDISCUSSIONREFERENCES |
|---|
cerebral blood flow; brain stimulation; neurovascular coupling; cerebral metabolism
In addition to clinical applications, noninvasive studies of dynamic CA, as afforded by Doppler ultrasound, are ideally suited to improve knowledge of cerebrovascular physiology in humans. Previous studies have shown that hypercapnia can reduce the efficiency of dynamic CA, whereas hypocapnia has the reverse effect, similar to the effects of arterial PCO2 on "static" autoregulation (1, 11, 26, 34). The influences of exercise (5, 23), posture (7, 8), intracranial pressure (30), body temperature (9), aging (6, 40), respiratory frequency (10), and autonomic nervous activity (3, 29, 42) have also been investigated. However, we are not aware of studies assessing the effects of cognitive and/or sensorimotor stimulation on dynamic CA. Activation of the brain increases the demand for O2 and CBF, thus shifting the static autoregulatory curve, but the corresponding effects on dynamic CA are difficult to predict. The increased sympathetic stimulation and hypocapnia that usually take place during cognitive and/or sensorimotor stimulation (36, 37) might suggest that the efficiency of dynamic CA could improve during brain activation paradigms (1, 4, 11, 42). To test this hypothesis, we described the ABP-CBFV dynamic relation by transfer function analysis (2, 3, 5, 10–14, 16–18, 22, 23, 25–27, 29, 30, 33, 35, 39, 41, 42) in a group of young healthy volunteers at rest and during repeated cognitive and motor tasks that are known to induce significant lateralization of hemispheric responses.
|
METHODS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
70% for right-handedness, as assessed by the Edinburgh handedness inventory (24), were accepted into the study. Fifteen healthy 21- to 43-yr-old subjects were recruited. The subjects avoided alcohol, nicotine, and caffeine-containing products for 24 h before arrival at a laboratory in which temperature (23°C) and lighting were controlled. The middle cerebral arteries (MCA) were insonated bilaterally using 2-MHz TCD ultrasound (model QVL-120, SciMed). The transducer probes were placed over the temporal bone window and locked in position by a head frame purposely designed for TCD ultrasound recordings. An ECG signal was monitored using three surface chest leads. End-tidal PCO2 (PETCO2) was measured via a closely fitting mask and an infrared capnograph (Capnogard, Novametrix). ABP was recorded noninvasively by finger photoplethysmography (Finapres 2300, Ohmeda) with the hand at heart level. All physiological signals were continuously recorded onto digital audiotape (PC-108M, Sony).
The data collection protocol has been described previously (21). After a 10-min baseline recording at rest, the subjects performed two different series of brain activation tasks, each lasting 10 min. Measurements were performed with subjects in the seated position, with the legs placed under a bench and the hand used to perform the tasks (i.e., the "active hand") resting on the bench. After random assignment, the ABP transducer was placed on the middle finger of the nonactive hand, the servo-correcting mechanism of the Finapres cuff was turned off, and the ABP calibration was recorded. Subjects were asked to breathe normally and relax with eyes closed during a 10-min baseline recording period. At the end of the recording period, the servo-correcting mechanism was switched on and then off, and a new calibration was performed before the first paradigm was started. The Finapres transducer was changed to the other hand, and a similar procedure was adopted to record another 10-min baseline segment of data before the second paradigm was performed.
The paradigm controlled by the right hand involved the random presentation of a letter on a computer screen for 30 s. During this time, subjects were asked to write as many words as possible beginning with the letter displayed on the screen. Subjects were asked to stop writing when the letter disappeared and to relax for the following 30 s while waiting for the next letter. The 10 letters selected correspond to the 10 most common initial letters in the English language. The paradigm controlled by the left hand was a simplified three-dimensional puzzle (37) using multicolored building blocks of different shapes. Ten different puzzles had been previously constructed and were displayed in random order on the computer screen for 30 s. Using their left hands, subjects were to select and pick up the individual blocks and assemble the puzzle as a two-dimensional pattern on the bench top. Subjects stopped as soon as the puzzle photograph disappeared from the screen and were asked to relax for the next 30 s before the next puzzle was displayed. The total duration of each paradigm was 10 min. A pulse of synchronism, indicating the beginning and end of each activation task, was also recorded on digital audiotape.
Data analysis. Data recorded on digital audiotape were downloaded onto a microcomputer in real time. A fast Fourier transform (FFT) was used to extract the maximum-frequency velocity envelope, with temporal resolution of 5 ms. The ABP, ECG, PETCO2, and stimulus marker signals were sampled at a rate of 200 samples/s, and ABP was calibrated at the start of each recording. All signals were visually inspected to identify artifacts or noise, and narrow spikes were removed by linear interpolation. The two CBFV signals were subjected to a median filter with a window width of five samples, and all signals were low-pass filtered by a zero-phase Butterworth filter with a cutoff frequency of 20 Hz.
The beginning and end of each cardiac cycle were detected on the ECG, and mean beat-to-beat values were calculated for the two CBFV channels, ABP, and heart rate. The end-tidal position was detected in the capnographic signal, and linear interpolation was used to obtain estimates of PETCO2 synchronized to the end of each cardiac cycle. All beat-to-beat estimates were interpolated using a third-order polynomial and resampled at 0.2-s intervals to generate a time series with a uniform time base. Suffixes R and L are used with abbreviations to denote right- and left-side variables, respectively.
Transfer function analysis was adopted to quantify the dynamic relation between ABP (input) and CBFV (output). Segments with 256 samples (51.2 s) were extracted, synchronized with the beginning of each activation task. Each segment was multiplied by a 20% cosine-tapered window after removal of its mean value. For the baseline recordings, 10 sequential segments, each with 256 samples, were extracted without superposition (2). Auto- and cross-spectral estimates were obtained by averaging the FFTs of the 10 sequential tasks for each paradigm or baseline recording.
The transfer function was calculated as the ratio of the smoothed cross-spectra [GPV(f)] to the autospectra of ABP [GPP(f)]
 | (1) |
 | (2) |
 | (3) |
, a separate spectral analysis was performed by delaying the CBFV signal by 2 s. This reduced the amplitude of the phase components, thus avoiding the phenomenon of "wraparound" (2). Subsequently, phase values were corrected using the corresponding linear phase shift 2fT, where T = 2 s and f is the frequency of each harmonic (2).
The squared coherence function, which represents the fraction of output power that can be linearly explained by the input power, was estimated as follows (2)
 | (4) |
Statistics. Repeated-measures ANOVA with four levels was used to test for changes in CBFV, ABP, heart rate, and PETCO2 between the four recording conditions (2 baseline and 2 different test activations). Post hoc comparisons were performed with Scheffé's test. Similarly, the presence of nonstationarity along repeated activation tasks was tested by means of a 10-level repeated-measures ANOVA for the standard deviation of CBFVR, CBFVL, and ABP using the 10 different segments of data extracted to estimate transfer function parameters for each of the 4 recording situations. Testing for changes in mean values was not necessary, inasmuch as the mean was normally removed from each data segment (256 samples) before estimation of the auto- and cross-spectra. Change from rest (OFF phase) to activation (ON phase) was tested with Wilcoxon's matched pairs test. For transfer function and power spectral estimates, testing was performed on the averaged values for harmonics below 0.05 Hz, where dynamic CA is usually assumed to be active (20, 26, 30, 41). For the CBFV step response, differences were tested using averaged values from 9–12 s into the response. Statistical significance was set at P < 0.05 for all tests.
|
RESULTS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
Brain activation with the word and puzzle paradigms induced very significant changes in mean CBFVR, CBFVL, ABP, and heart rate, but not PETCO2 (Table 1). Post hoc comparisons showed no significant differences between the two baseline conditions or between the two activation paradigms: all differences were due to changes from baseline to activation.
|
The highly significant changes from baseline to activation for ABP, CBFVR, and CBFVL were confirmed by observing the change in power spectra from baseline to activation (Figs. 1 and 2 and Table 2). The average power CBFVR and CBFVL, at <0.05 Hz [excluding the DC (0-frequency) term], was significantly greater during activation than at baseline (Fig. 2 and Table 2), but the ABP spectral power was increased only during the word tasks (Fig. 1 and Table 2). PETCO2 spectral power at <0.05 Hz was also significantly increased during both paradigms (Table 2). The coherence function was higher during activation than at baseline for right and left MCA recordings (Fig. 3), but statistically significant differences were obtained only for the right MCA during the puzzle tasks and the left MCA for the word paradigm (Table 2). The amplitude of the frequency response (Eq. 2) was also greater during activation than at rest, and the difference was significant for both hemispheres during the puzzle paradigm and for the left MCA for the word tasks (Table 2). Significant reductions in phase (Eq. 3) were obtained for the puzzle task (right and left MCAs) and word paradigm (right MCA), although the left MCA also showed a reduced mean phase during activation in relation to baseline (Table 2).
|
|
|
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
Transfer function analysis of the relation between spontaneous or induced fluctuations in ABP and the corresponding oscillations in CBFV has been used by many investigators as a tool to assess the effectiveness of dynamic cerebral pressure autoregulation in humans. According to this model, a deterioration of dynamic CA would be reflected by changes in the main frequency and time domain functions that characterize the pressure-flow dynamic relation. In normal subjects, the coherence function tends to be low (<0.5) for frequencies below 0.1 Hz, reflecting the relative independence of changes in CBFV from ABP fluctuations in this frequency range, but other factors can lead to fluctuations in CBFV (see below). Worsening of dynamic CA should lead to increases in coherence as CBFV tends to follow ABP more closely (12). The gain, or amplitude frequency response, is normally reduced in the frequency range DC to 0.1 Hz, in relation to corresponding values at higher frequencies. This quantity also tends to increase with impairment of dynamic CA, again reflecting the increased influence of ABP on CBFV fluctuations. One key feature of dynamic CA in normal subjects is the phase lead of CBFV in relation to ABP, resulting from the feedback effects of adjustments in cerebrovascular resistance (CVR) (15). With deterioration of dynamic CA, the phase tends to zero, thus reflecting the failure of CVR to respond to changes in ABP. Finally, the time domain functions, such as the CBFV impulse and step responses, can also give an indication of dynamic CA efficiency. The latter is ideally suited for this purpose, because it can provide a result analogous to the typical response to a sudden change in ABP induced by the thigh cuff test (1). Under normal conditions, a step change in ABP produces a sudden change in CBFV, which then recovers to its original level within a few seconds. As dynamic CA becomes less efficient, the return of CBFV slows and often does not reach the original baseline level. Clinical conditions where CA could be expected to be impaired, such as carotid artery disease (14, 35), subarachnoid hemorrhage (18), severe head injury (22, 33), ischemic stroke (17), MCA stenosis (13), autonomic failure (3, 42), malignant hypertension (16), and neonatal prematurity (25, 39), support the predicted changes in coherence, gain, phase, and CBFV step responses described above. Similar behavior was also observed during hypercapnia, which is known to reduce the efficiency of static and dynamic CA (1, 11, 26, 34).
In the light of the accumulated evidence regarding the interpretation of transfer function estimates of the ABP-CBFV relation mentioned above, the results in Table 2 and Figs. 3 and 4 suggest that sustained mental activation can reduce the efficiency of dynamic CA, thus implying rejection of our original hypothesis. Despite this clear-cut result, other interpretations are possible and should be considered. Mental activation is known to lead to increases in CBF coupled to increased metabolic demand. Consequently, in addition to the CBFV component linked to fluctuations in ABP, during activation there will be an additional metabolic component of flow that is independent of ABP. By relating both flow components to the ABP input, it is possible that the metabolic component will be reflected as increases in gain, reductions in phase, and, mainly, an increased and delayed tail on the CBFV step responses (Fig. 4). On the other hand, this argument cannot explain the changes observed in the coherence function (Fig. 3). If the metabolic component is independent of ABP, it should lead to a reduction in coherence, rather than the increases we have observed in the frequency range where dynamic CA is known to be active (Fig. 3).
Worsening of dynamic CA should also be expected if mental activation had led to hypercapnia. During the puzzle tasks, there was a small but nonsignificant increase in PETCO2 (Table 1), and PETCO2 power increased during both tasks (Table 2). The increase in power indicates larger fluctuations, but, on average, PETCO2 decreased during activation and increased again during the interactivation phase (21). Other investigators showed that mental activation can lead to hypocapnia as a result of hyperventilation (10, 34, 36, 37). The dynamic response of CBFV to changes in PETCO2 has been modeled by transfer function analysis (26), and more complex, multivariate models (11, 20, 28) might account for the influences of PETCO2 on estimates of dynamic CA. If mental activation had led to significant hypercapnia, inclusion of PETCO2 in a multivariate model could possibly explain the deterioration of dynamic CA, as shown by previous investigators (1, 11, 26) as an effect of CO2 and not mental activation. However, because PETCO2 actually decreases during activation, it is likely that a multivariate model will show that the net contribution of CO2 is to attenuate the observed deterioration of dynamic CA derived from the ABP-CBFV relation, meaning that if PETCO2 had remained constant, even greater worsening of dynamic CA would have been observed.
Misleading results from classical linear transfer function analysis could be obtained if the relation between ABP and CBFV was strongly nonlinear (2). Rigorously, CA is a nonlinear phenomenon due to the modulation of CVR (15). However, it has been demonstrated that, for small changes in ABP and CBFV, as observed during spontaneous fluctuations, the linear model provides an acceptable approximation (25, 41). The nonlinear nature of the system would be more likely to be manifested during larger excursions in ABP and CBFV, as induced by thigh cuff or Valsalva maneuvers. Even in these situations, the suggestion of marked nonlinearities could not be demonstrated (29, 41). More recently, Mitsis et al. (20) studied the joint contribution of ABP and PCO2 as determinants of CBFV spontaneous variability and concluded that there are important nonlinearities, mainly from PCO2 and its interaction with ABP, whereas the contribution of ABP to CBFV was mostly linear. In our study, there were no significant changes in mean PETCO2 (Table 1), but it might be possible that larger fluctuations in PCO2 during activation (Table 2) could have interacted with ABP to produce the apparent worsening of dynamic CA reflected by the changes in coherence, gain, phase, and CBFV step responses. We have shown that activation led to significant increases in mean values of ABP and CBFV (Table 1), as well as in their spectral powers (Table 2). Nevertheless, from the increased power ratios, it is possible to estimate that increases in signal amplitude were less than twice the increase observed during baseline and, hence, less than the changes induced by thigh cuff and Valsalva maneuvers. On the other hand, it is possible that synergistic interaction of myogenic and metabolic activity during mental activation could change the approximately linear nature of the ABP-CBFV relation at rest ("baseline") to more pronounced nonlinear behavior during stimulation. If this were the case, one would expect a further reduction in the low coherence observed in the low-frequency band (Fig. 3), because this function only reflects the amount of output power linearly related to the input signal. The observation that coherence has increased in the low-frequency range gives an indication that mental activation not only increased the coupling between ABP and CBFV but did so in a linear fashion.
One limitation of our study is the use of Doppler ultrasound to obtain estimates of CBF. CBFV can be used as a reliable surrogate of CBF only if the diameter of the insonated artery remains approximately constant. Mental activation could lead to constriction of the MCA due to increased sympathetic stimulation. In this case, CBFV would give an overestimation of changes in CBF. The disproportional increase in CBFV could explain the observed increases in gain (Table 2) but should not have an influence on the coherence and phase responses, because these functions are not sensitive to amplitude changes. The CBFV step response should not be affected, because it was normalized for amplitude (Fig. 4). The use of a relatively narrow (51.2-s) FFT window can also be seen as a limitation due to the correspondingly higher coefficient of variation of spectral estimates in relation to the longer window durations normally used in transfer function analyses of dynamic CA and the reduced frequency resolution (2, 5, 12, 25, 41). We repeated our analysis with longer windows (512 samples or 102.4 s) and obtained almost identical values for the phase distributions and step responses but higher values for the coherence and gain at very low frequencies (<0.05 Hz) during activation. The problem with >60-s windows is that they span more than one cycle of activation, thus leading to artificially high coherences and gain, which have more to do with the on-off transition of activation than with the ABP-CBFV relation during activation. For this reason, we opted for a window that includes only one cycle of activation to avoid the possibility of artifacts and compensated for the higher variability of spectral estimates by averaging the cross- and autospectra for 10 sequential activation cycles.
One other important factor to consider in transfer function analyses is the potential presence of nonstationarities in the data, which could lead to distorted results (2). Two different aspects need to be considered here. First, spectral estimates (Eqs. 1 and 4) should only be extracted for segments of data that are stationary. Because mean values were removed from the ABP and CBFV time series before application of the FFT, we only need to worry about the second moment, i.e., the variance or standard deviation of the signals. The repeated-measures ANOVA confirmed that there were no significant differences in power (or variance) for the 10 segments of data used to calculate spectral estimates for baseline and activation conditions; consequently, each situation can be regarded as stationary. The second aspect that needs attention is the change from rest (or baseline) to activation. As shown by the results in Table 2 and Figs. 1 and 2, mental activation leads to significant increases in ABP and CBFV power, with the obvious implication that activation represents a departure from the stationary conditions prevailing during the rest phase, and vice versa. Therefore, activation represents a shift from one stationary state to another, and, as long as spectral estimates are derived separately for each condition, there is no breach of the stationarity requirements of classical transfer function analysis. This situation is similar to recent studies of the effects of exercise on dynamic CA (5, 23).
Our present discussion emphasizes the need for further work to clarify whether mental activation leads to depression of dynamic CA or whether transfer function analysis of the ABP-CBFV relation can be a misleading tool when CBFV is modulated by other physiological variables. In previous studies, we observed considerable longitudinal variability of indexes of dynamic CA (31, 32), which could be partially explained by fluctuations in mental activity during relatively long recordings lasting up to 10 min. Similar findings were recently reported by Mitsis et al. (20). We and others have also shown that it is possible to extend classical bispectral analysis to include other determinants of CBF, such as PCO2, by using time-domain multivariate autoregressive modeling (11, 20, 28). A similar approach, where mental activation is included as an independent variable, might help to clarify the extent to which pressure autoregulation is truly affected by cognitive and/or sensorimotor stimulation.
The main conclusion of our study is that classical transfer function analysis of the ABP-CBFV relation suggests that dynamic CA is depressed during mental activation tasks. Whether this result reflects a true physiological phenomenon or is an artifact due to limitations in bispectral analysis, there is a clear message that either possibility can influence other studies of human dynamic CA where there might be variable levels of mental activity. In this respect, particular care should be taken in studies of the effects of exercise or the use of hand grip to stimulate changes in ABP. Finally, clinical applications of dynamic CA should take into account potential interactions between changing levels of mental activity and an already compromised CA.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
|
REFERENCES |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  E. L. Sammons, N. J. Samani, S. M. Smith, W. E. Rathbone, S. Bentley, J. F. Potter, and R. B. Panerai Influence of noninvasive peripheral arterial blood pressure measurements on assessment of dynamic cerebral autoregulation J Appl Physiol, July 1, 2007; 103(1): 369 - 375. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 B. Kolb, D. L. Rotella, and H. M. Stauss Frequency response characteristics of cerebral blood flow autoregulation in rats Am J Physiol Heart Circ Physiol, January 1, 2007; 292(1): H432 - H438. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| Visit Other APS Journals Online |
