Methods: On averaging continuous raw data

We now have the luxury (compared to the good old days) of continuously monitoring data during a study protocol. For example, one can acquire continuous beat-to-beat measurements of middle cerebral blood flow velocity (MCA V) during 5, 30, 180 minutes or even more.

When you start a new study, you may want to analyze your raw data:

1) using the same method described in that interesting paper you just read or,

2) as it has always been done in your lab.

So, at the end of data collection, you sit in front of your raw data which now need to be averaged. What’s your game plan ?

If you compare two 10-minute conditions (baseline and drug infusion; under steady-state for the last 5 min), you get your data averaged over the last 5 seconds of steady-state? 15 seconds ? 60 seconds? 180 seconds ? Why? Also, is it protocol-depend? Why?

Going for the jugular: the end

Regular readers will remember my research interests. Thanks to good friends, it has now become essential for my research projects evaluating the influence of certain vasopressors on brain oxygenation to invasively measure the levels of oxygen in the brain (specifically in the blood coming from the internal jugular vein). Indeed, a recent paper from my Danish colleagues suggested that changes in brain oxygenation (measured non-invasively by near-infrared spectroscopy), following manipulation of blood pressure by norepinephrine, could be partly influenced by changes in skin perfusion. I had the opportunity to measure jugular venous oxygen saturation in healthy volunteers and patients during my postdoc (in Copenhagen with that same group) and this is definitely a “must” for the kind of research project I am working on.

In front of this evidence, I decided to make an amendment to my current vasopressor study in order to add jugular venous oxygen saturation to existing measures. But remember, I am not in Copenhagen anymore…I didn’t know if it would be difficult for this amendment to be accepted.

I thus have presented my amendment to our Scientific committee first and soon received a positive answer. It was pretty scientifically obvious that we needed that measure. Then, we have (for that specific meeting, I have asked the anesthesiologist who will insert these lines to be present) tried to convince the Ethics committee…

Well, I am happy to inform you that blood samples from the internal jugular vein will be taken in that vasopressor study! Yay! As I have already mentioned,  I am aware of only one lab in Canada that can insert catheters into the internal jugular vein in human volunteers. For the moment, I will be mostly interested in jugular venous oxygen saturation. However, I will now have the possibility of studying cerebral metabolism.

For those wondering what a catheter into the internal jugular vein looks like, well here it is (the catheter is advanced until the black arrow):


I am looking forward to seeing the results !!!

Methods: Monitoring cerebral blood flow velocity with transcranial Doppler: hand-held vs. frame-held probe

When I was introduced to cerebrovascular physiology a couple years ago, the first piece of equipment I’ve worked with was transcranial Doppler. After having learned the technique to search flow velocity in the middle cerebral artery, the following was pretty clear to me: 1) a good quality Doppler signal was critical for subsequent interpretation and, 2) you first use a hand-held probe to search for the artery of interest and then switch to a frame-held probe (see figure below) to monitor, usually for a continuous period of time (for example, continuous infusion of vasopressors or endurance exercise), blood flow velocity.


That frame-held probe is appealing since, among other things, you need to keep the same angle of insonation throughout the study protocol in order to have reliable blood flow velocity data (because there is no way to know that angle of insonation with transcranial Doppler) especially in situations where the head is moving during data acquisition.

So, in the beginning of my training, the utilization of a frame-held probe was the way to go in cerebrovascular research since I had only worked with healthy volunteers/patients without head injuries…but what about these individuals for whom these head frames are most likely not tolerated ? Could we use hand-held probes and still be confident in our data acquisition ? What about stability and reliability of hand-held probes?

The main objective of an interesting study recently published by Saeed et al. was to compare cerebral blood flow velocity and cerebral autoregulation using a hand-held vs. frame-held probe.

Cerebral blood flow velocity (in the middle cerebral artery) was thus monitored with a frame-held probe or held in position by one investigator for 5 minutes (for each method) in 11 heatlhy volunteer. Hemodynamics was also monitored and cerebral autoregulation calculated for each period. These measurements were repeating in a subsequent visit. A total of 3 frame-held and 2 hand-held measurements were performed on both left and right sides at each visit.

The main findings of this study were:

1) no significant differences in cerebral blood flow velocity and cerebral autoregulation between the methods;
2) no significant difference between repeated measures on separate days for both cerebral blood flow velocity and cerebral autoregulation.
As highlighted by the authors:

(…) this is the first study to report that HH measurements of CBFv and ARI estimates are not statistically different from FH, which has important implications for bedside estimates of CBFv and ARI, particularly in traumatic brain injury (TBI) patients and neonates, where CBFv and ARI measurements form an important part of clinical assessment.

(HH: Hand-held; CBFv: cerebral blood flow velocity; ARI: cerebral autoregulation; FH: frame-held)

These results are very interesting and promising if one wants to investigate cerebral blood flow velocity with transcranial Doppler in patients who cannot tolerate a head frame. However, we have to note that a short period of recording was investigated in this study. It will be important to replicate this protocol with longer recording periods.


Saeed NP, Panerai RB, Robinson TG. Are hand-held TCD measurements acceptable for estimates of CBFv? Ultrasound Med Biol 2012 38(10):1839-44

Methods: Cerebral blood flow velocity; Curve fitting

A regular reader of this blog had an excellent idea. She suggested discussing methods related to cerebrovascular physiology. I will thus try to come up with posts related to specific methods used to study cerebral blood flow, cerebral oxygenation, cerebral autoregulation, etc.  This new section will also be used to answer your questions (to the best of my knowledge) regarding methods used in this research field. I don’t pretend to know everything. So if you have any comments to add to the discussion, it will be very appreciated!

In this first post, I am answering Jess Inskip’s questions.

Question: Why do some researchers describe mean CBFV as the arithmetic mean, while others use the integral under the envelope of the CBFV tracing? Surely the latter is more important, as it is the waveform that seems to be sensitive to perturbations rather than the raw SBF and DBF.

Answer: Mean cerebral blood flow velocity (Vmean) is measured using velocity associated with maximal frequencies of the Doppler shifts (envelope). Vmean can be used as an index of cerebral blood flow if we assume that: 1) maximal velocity is proportional to the mean velocity of the blood flow in the vessel; 2) the vessel’s cross-sectional area remains constant. Although Vmean in the MCA increases in parallel with the inflow of the internal carotid artery, with the initial slope index of the 133Xenon clearance-determined cerebral blood flow, and with regional cerebral blood flow measurements by positron emission tomography during exercise for example, some argue that changes in velocities associated with the maximal frequencies of the Doppler shift doesn’t necessarily reflect changes in flow. These authors use the intensity-weighted mean velocity (VIWM), which is based on the complete velocity spectrum. The latter represents a mean velocity averaged over the entire cross section of the blood vessel. However, while small changes in the angle of insonation of a cerebral artery is of no consequence for the maximal velocity, the average flow velocity is affected by the angle of insonation. So, other investigators such as Secher et al. (J Appl Physiol 2006) argue that changes in cerebral flow velocity during exercise are essentially lost when attention is not on Vmean.

Interesting reads:

Willie CK et al. J Neurosci Methods 196:221-237, 2011

Lohmann H et al. in Baumgartner RW (ed): Handbook on Neurovascular Ultrasound. Front Neurol Neurosci. Basel, Karger, 2006, vol 21, pp 1–10

Aaslid R et al. J Neurosurg 57:769-774, 1882

Poulin MJ et al. J Appl Physiol 86:1632-1637, 1999

Question: Is there a way to determine how to curve fit the relationship between PETCO2 and cerebral blood flow. Many seem to continue to define this as a linear relationship, yet other recent research groups have shown that this should be a sigmoidal relationship (even when these two groups seem to be working in the same PETCO2 ranges).

Answer: The utilization of the right model to study this relationship remains ambiguous to me. First, you need to choose the technique to evaluate the cerebrovascular reactivity to CO2.  Accordingy to Ainslie (see Interesting reading for the reference), you need to choose the method to evaluate cerebrovascular reactivity to CO2 according to your experimental question, and a clear choice of method to study it (steady-state changes in CO2 vs. modified rebreathing) is problematic until the physiology of cerebrovascular control is better understood.

I think that, for the moment, it is the same for the determination of the right model to study the relationship between PaCO2/PetCO2 and cerebral blood flow, since there exists a plethora of models. Indeed, you can model steady-state cerebrovascular CO2 reactivity using a least squares linear regression, a biasymptotic curve analysis;, a dynamic single-compartment model, an exponential function, etc. When using modified CO2 rebreathing, a lot of studies used linear regression although some investigators demonstrated a non-linear relationship between PaCO2 and cerebral blood flow.

So, it is true that several studies are quantifying cerebrovascular reactivity to CO2 using a linear regression. One simple reason is that more complex modeling of cerebrovascular reactivity to CO2 is too complicated for clinical use.  Also, more data points are needed with an exponential model vs. a linear model…Still, an interesting benefit with exponential models = more precise data fitting. This being acknowledged, a recent study suggests that the fitting strategy doesn’t seem to influence results…at least for the estimated PaCO2 range of interest (20-60 mmHg). So, is there a way to determine how to curve fit this relationship? It depends, most likely, on the aim of your study…

Question: Furthermore, if you are looking at CO2 reactivity at resting conditions and after perturbation X, what if one of the effects of the perturbation x is in fact altering the shape of the relationship between PETCO2 and CBF.

Answer: You have an excellent point. I don’t necessarily have the answer. What I can tell is that reactivity to CO2 is higher during hypercapnia vs. hypocapnia (potential mechanism: more important impact of vasodilator vs. vasoconstrictor mediators on vascular tone). In addition, factors such as hydration, caffeine, time of day, posture, etc. have an influence on the relationship between cerebral blood flow and reactivity to CO2. Finally, this relationship could also be affected by CO2-induced changes in arterial pressure.

Interesting read:

Ainslie P et al. Am J Physiol Regul Integr Comp Physio 296: R1473-R1495, 2009