CourseNeuronal Data Analyzer Lab

Lesson 1 · Laser Doppler flowmetry

0%

Lesson 1 · Laser Doppler flowmetry

The blood-flow response to whisker stimulation, with laser Doppler flowmetry

From the probe to a number you can defend in a paper. Play with each figure, answer the questions, and do the steps in the app.

  • 7parts
  • 60minutes
  • 7questions
  • 5numbers to check

No code. You need the toolbox and its demo files (before you start). Stop after any part: your progress stays in this browser.

Draft for review. Boxes marked From the lab will be written by the author from his own experience; until then they list the questions they will answer.

Start with Part 1

Part 1 of 7 · 10 min

What the probe measures

Neurons areactiveAstrocytes andvessel cells signalArterioleswidenMore red cellsflow: LDF riseswithin about a second of the stimulus
Active neurons need more blood: the flow over them rises within about a second19. This is neurovascular coupling6,7. Whisker stimulation drives it in the barrel cortex9,10.
tissuelaserinlightbackprobered cells movinghits still tissue:same frequencyhits a moving cell:frequency shifted(Doppler)
The probe turns the frequency shifts of the light into one number, the perfusion, in arbitrary perfusion units (PU)1–3. One spot, one trace over time.

Try it

Perfusion ≈ number of moving red cells × their speed2. Dilated capillaries hold more moving cells, and they move faster.

Check yourself · Question 1 of 7

The LDF signal rises when…

Relative, not absolute

stimulus100150200LDF (PU)Animal A180 → 225 PU+45 PU · +25%Animal B95 → 119 PU+24 PU · +25%time (s)
Raw PU change with the spot under the probe. Compare each animal's change from its own baseline, in PU and %4,5.

Check yourself · Question 2 of 7

Animal A has a baseline of 180 PU, animal B of 95 PU. What can you conclude?

Go deeperThe evidence, and why "cells × speed"

Against a gold standard (autoradiography) in rat cortex, absolute LDF values correlated poorly with flow (r = 0.54) but percentage changes correlated well (r = 0.91)4. A second study found the same (absolute r = 0.44) and that LDF tends to report larger increases than the reference5: treat large % increases as an upper estimate.

Light in tissue is scattered many times, so the Doppler shifts form a broad spectrum. Its first moment grows with the number of moving cells times their mean speed, as long as few photons meet more than one moving cell2. Over a large vessel that fails: one more reason to avoid them.

A typical response to a few seconds of whisker stimulation in an anaesthetised rat is +25–31%10. It is smaller during cortical arousal12 and under strong hypercapnia18, and stimuli longer than about 2 s no longer add up simply11.

Part 2 of 7 · 10 min

Recording well

The analysis cannot fix a missing baseline, overlapping trials or too few animals. Decide them at the bench.

TriggerLDF≥ 60 s baselinejudge drift30 s apartback to baseline first5 s stimulus,same every time060120180240300 s
Record the stimulus trigger in the same file as the LDF: every later step finds the trials from it9.

Try it

Move the spacing below the length of the response: trial 2 starts on a raised baseline and its increase shrinks.

Check yourself · Question 3 of 7

A colleague stimulates every 8 s to collect more trials. The response lasts about 12 s. What goes wrong?

Stable probeIn a micromanipulator, perpendicular to the surface.
No large vesselIt dominates the signal. Baselines outside 30–600 PU are flagged.
Thinned skull or windowLess invasive than removing the bone9.
Stable animalAnaesthesia, temperature, blood gases change the response12,18.

From the lab

To be written by the author from his own experience. It will answer:

  • Which anaesthesia, and how long after induction before recording?
  • What is monitored (temperature, blood gases, blood pressure, breathing), and which ranges are accepted?
  • Thinned skull or craniotomy? Placing the probe over the barrel field and away from surface vessels.
  • Which sampling rate, and why it is enough for a response that lasts seconds?
  • The stimulus: frequency, duration, intensity; trials per animal; rest blocks.
  • How you tell a recording is good during the experiment; when you stop or discard an animal; the most common mistakes of new lab members.

Part 3 of 7 · 10 min In the app

Look at the raw data first

Try it

A raw trace is a sum of parts. Switch them off one by one, and zoom in to see the 6 Hz heartbeat. The slow 0.13 Hz wave is vasomotion: real physiology, not a fault.

In the app

  1. Blood flow → Laser Doppler flowmetry → 1 Extract, Load file..., demo_ldf_export.mat (simulated: 300 s at 1000 Hz).
  2. Count the stimulus pulses: 9. Read the baseline: about 120 PU. See a rise after every pulse.
Extract LDF with the demo recording: nine stimulus pulses on top, the LDF trace below with a rise after each pulse
What you should see. Demo data.

What faults look like

Driftthe baselinecreeps upMovementa jump bigger thanany responseSignal at 0the probe lostcontact
Drift: check the animal's state, measure each trial against its own baseline. Movement: exclude the trial, by a rule written in advance. Zero: the probe lifted; exclude that part.

In the app

Open demo_ldf_faults.mat in 2 LDF Processing, segment it with the Part 4 settings, open the Checks tab and click a row.

The Checks tab of LDF Processing on the faults recording: baseline drift to check, movement artefacts and signal range as warnings
The toolbox finds all three. Demo data.

Part 4 of 7 · 20 min In the app

Processing, step by step

  1. Crop20–280 s1 Extract
  2. Downsample1000 → 100 Hz2 Processing
  3. Filterlow-pass 1 Hz2 Processing
  4. Cut trials−5 to +20 s2 Processing
  5. Averagerelative to baseline3 Average

1 · Crop

Keep the experiment, drop the setup at the start and the end. Crop by the protocol, never by how a response looks.

In the app · 1 Extract

Start 20, End 280, Crop to range, Save cropped data.... The first pulse is now at 10 s: its 10 s before are its baseline.

Extract LDF after cropping 20 to 280 s
After cropping. Demo data.

2 · Downsample

Try it

A sampled signal shows only frequencies below half its rate (the Nyquist frequency). Faster ones fold back as false slow waves (aliasing), unless they are filtered out first. The toolbox always filters first.

In the app · 2 LDF Processing

Load the cropped file, Settings..., Downsample 10x (100 Hz: plenty for a response that lasts seconds).

3 · Filter

Try it

Find the low-pass cutoffs that remove the heartbeat and keep the peak. Then switch to high-pass: a 5 s response is a slow change, so a high-pass eats it14.

In the app · same dialog

Low-pass, Butterworth, High cutoff 1 Hz, Order 4, Apply. The Filter response tab shows what is kept.

LDF Processing after downsampling and a 1 Hz low-pass filter
Heartbeat gone, responses unchanged. Demo data.
Go deeperZero-phase, and choosing a cutoff

Every filter delays what passes through it. The toolbox runs it forwards, then backwards (zero-phase): the delays cancel and the peak stays in place. A Butterworth filter is flat in the band it keeps13, so it adds no ripples.

Keep the fastest change you need: the rise of the response, 1–2 s, carried below about 1 Hz. The Filter check flags low-pass cutoffs below 0.5 Hz and high-pass cutoffs above 0.1 Hz. A higher order cuts more sharply but rings more on steps.

4 · Cut trials

file ends (260 s)123456789Each window: 5 s before the onset to 20 s after9: only 10 safter it:skipped8 of 9 trials fit
A trial that does not fit is skipped, never padded. The status line and the Checks tab say so.

In the app · 2 LDF Processing

Threshold 2.5 (half of the 5 V pulse), pre 5 s, post 20 s, minimum interval 10 s, Segment trials, Save trials....

The Trials tab: 8 trials from −5 to 20 s around each onset and their mean
8 trials and their mean ± SD. Demo data.
Go deeperThe three settings

The threshold sits between the trigger's off and on levels. The minimum interval stops a pulse train, or noise on the trigger, from counting as several stimuli. The window covers a baseline and the whole response without reaching the next stimulus.

5 · Average

Try it

Add trials: the response is the same in each, vasomotion and noise are not, so they cancel. 4 times the trials, half the uncertainty.

In the app · 3 Average

Add files... (your trials), tick Relative to baseline (each trial minus its own mean before 0 s), Plot grand average. Write down the numbers under the button.

Average LDF Viewer: all trials and the grand average relative to baseline, with baseline, peak increase and time to peak under the button
The grand average and its numbers. Demo data.

Check yourself · Question 4 of 7

The band around the mean is ± SD. What does it tell you?

Part 5 of 7 · 8 min

Numbers, and what n is

stimulus120140160-505101520time from stimulus onset (s)Baseline: mean before 0 sPeak increase (PU, % of baseline)Time to peakback to baseline
Compare the % between animals, not the raw baseline. Time to peak is known only to within a sample, and is noisy near a flat peak.
Controlrat 1rat 2rat 3rat 4Treatedrat 1rat 2rat 3rat 410 trials per rat → one average per rat → n = 4 per group, not 40
Trials of one animal are not independent. Counting them as n is pseudoreplication, and makes chance look like an effect16. The Signal Characterization window compares animals and checks the test.

Check yourself · Question 5 of 7

You recorded 10 trials in each of 4 rats per group. What is n for comparing the groups?

Error bars: say which. SD shows the spread; SE and confidence intervals show how precisely a mean is known15.

From the lab

To be written by the author from his own experience. It will answer:

  • How many animals per group do you plan for this kind of study, and how?
  • Which exclusion rules (trials, animals) do you write down before the analysis?

Part 6 of 7 · 5 min In the app

Reporting

Could a reader redo your analysis? Click Methods text... in each window for a draft with every setting, then tick what it covers and add the rest yourself.

Part 7 of 7 · 10 min

Apply it

A. Your numbers

Type what the Average window showed after Part 4. They are compared with a careful analysis of the same file, and stay in this browser.

Check your numbers

NumberYour answer

Answer keyWhat was true (the simulation)

Baseline about 120 PU with slow drift, vasomotion at 0.13 Hz, heartbeat at 6 Hz and noise; every stimulus adds +30 PU (25%) peaking 4 s after it starts, back after about 12 s. Your 8-trial average comes close, not exactly: that is why you are compared with a careful analysis of the same file.

B. A recording with problems

Process demo_ldf_faults.mat the same way. Write down what you see in the Signals tab before you open Checks.

Check yourself · Question 6 of 7

One movement artefact falls inside trial 3. What do you do?

Check yourself · Question 7 of 7

The baseline drifts by about +5% per minute. Is the response still usable?

References

  1. Stern MD (1975). In vivo evaluation of microcirculation by coherent light scattering. Nature 254:56–58. doi:10.1038/254056a0
  2. Bonner R, Nossal R (1981). Model for laser Doppler measurements of blood flow in tissue. Appl Opt 20(12):2097–2107. doi:10.1364/AO.20.002097
  3. Leahy MJ, de Mul FF, Nilsson GE, Maniewski R (1999). Principles and practice of the laser-Doppler perfusion technique. Technol Health Care 7(2–3):143–162. PubMed 10463304
  4. Dirnagl U, Kaplan B, Jacewicz M, Pulsinelli W (1989). Continuous measurement of cerebral cortical blood flow by laser-Doppler flowmetry in a rat stroke model. J Cereb Blood Flow Metab 9(5):589–596. doi:10.1038/jcbfm.1989.84
  5. Fabricius M, Lauritzen M (1996). Laser-Doppler evaluation of rat brain microcirculation: comparison with the [14C]-iodoantipyrine method suggests discordance during cerebral blood flow increases. J Cereb Blood Flow Metab 16(1):156–161. doi:10.1097/00004647-199601000-00018
  6. Attwell D, Buchan AM, Charpak S, Lauritzen M, MacVicar BA, Newman EA (2010). Glial and neuronal control of brain blood flow. Nature 468:232–243. doi:10.1038/nature09613
  7. Iadecola C (2017). The neurovascular unit coming of age: a journey through neurovascular coupling in health and disease. Neuron 96(1):17–42. doi:10.1016/j.neuron.2017.07.030
  8. Lauritzen M (2005). Reading vascular changes in brain imaging: is dendritic calcium the key? Nat Rev Neurosci 6(1):77–85. doi:10.1038/nrn1589
  9. Gerrits RJ, Stein EA, Greene AS (1998). Laser-Doppler flowmetry utilizing a thinned skull cranial window preparation and automated stimulation. Brain Res Protoc 3(1):14–21. doi:10.1016/s1385-299x(98)00016-6
  10. Peng X, Carhuapoma JR, Bhardwaj A, Alkayed NJ, Falck JR, Harder DR, Traystman RJ, Koehler RC (2002). Suppression of cortical functional hyperemia to vibrissal stimulation in the rat by epoxygenase inhibitors. Am J Physiol Heart Circ Physiol 283(5):H2029–H2037. doi:10.1152/ajpheart.01130.2000
  11. Martindale J, Berwick J, Martin C, Kong Y, Zheng Y, Mayhew J (2005). Long duration stimuli and nonlinearities in the neural–haemodynamic coupling. J Cereb Blood Flow Metab 25(5):651–661. doi:10.1038/sj.jcbfm.9600060
  12. Jones M, Devonshire IM, Berwick J, Martin C, Redgrave P, Mayhew J (2008). Altered neurovascular coupling during information-processing states. Eur J Neurosci 27(10):2758–2772. doi:10.1111/j.1460-9568.2008.06212.x
  13. Butterworth S (1930). On the theory of filter amplifiers. Experimental Wireless and the Wireless Engineer 7:536–541.
  14. Widmann A, Schröger E, Maess B (2015). Digital filter design for electrophysiological data – a practical approach. J Neurosci Methods 250:34–46. doi:10.1016/j.jneumeth.2014.08.002
  15. Cumming G, Fidler F, Vaux DL (2007). Error bars in experimental biology. J Cell Biol 177(1):7–11. doi:10.1083/jcb.200611141
  16. Lazic SE (2010). The problem of pseudoreplication in neuroscientific studies: is it affecting your analysis? BMC Neurosci 11:5. doi:10.1186/1471-2202-11-5
  17. Percie du Sert N, Hurst V, Ahluwalia A, et al. (2020). The ARRIVE guidelines 2.0: updated guidelines for reporting animal research. PLoS Biol 18(7):e3000410. doi:10.1371/journal.pbio.3000410
  18. Jones M, Berwick J, Hewson-Stoate N, Gias C, Mayhew J (2005). The effect of hypercapnia on the neural and hemodynamic responses to somatosensory stimulation. NeuroImage 27(3):609–623. doi:10.1016/j.neuroimage.2005.04.036
  19. Weber B, Burger C, Wyss MT, von Schulthess GK, Scheffold F, Buck A (2004). Optical imaging of the spatiotemporal dynamics of cerebral blood flow and oxidative metabolism in the rat barrel cortex. Eur J Neurosci 20(10):2664–2670. doi:10.1111/j.1460-9568.2004.03735.x

Each reference was checked against PubMed (Butterworth 1930 against the journal) before it was cited.

How was this lesson?

Something unclear or missing, or an idea for the next lesson? Tell us on GitHub (a free GitHub account is needed; what you write there is public).

All lessons