Brain, Respiration, and Cardiac Causalities in Anæsthesia · 2016-04-15 · Brain, Respiration, and Cardiac Causalities in Anæsthesia Origins, basis, rationale and science Peter

Brain, Respiration, and CardiacCausalities in Anæsthesia

Origins, basis, rationale and science

Peter V. E. McClintock1, Martin Hasler2, Juergen Kurths3,Per Kvandal4, Svein Landsverk4,

Dmitri G. Luchinsky1, Milan Paluš5, Arkady Pikovsky3,Zvezdan Pirtošek6, Johan C. Ræder4,

Samo Ribaric7, Michael Rosenblum3, Andrew F. Smith8,Aneta Stefanovska1,9, and Niels Wessel3

1Lancaster, 2Lausanne, 3Potsdam, 4Oslo, 5Prague, 6Ljubljana UMC,7Ljubljana FM, 8Lancaster RLI, 9Llubljana FEE

Lancaster – 12 April 2016BRACCIA Consortium The BRACCIA enterprise

Outline

1 IntroductionPhysiological oscilllationsAnæsthesiaData analysis methods

2 The BRACCIA enterpriseConsortiumResearch programme

3 ConclusionSummary

How do the oscillations and their couplings changein anæsthesia? Might these changes provide anovel basis for measuring depth of anæsthesia?

BRACCIA Consortium The BRACCIA enterprise

IntroductionThe BRACCIA enterprise

Conclusion

Physiological oscilllationsAnæsthesiaData analysis methods

Averaged wavelet spectra

0.18 Hz

1.1 Hz

0.18 Hz

0.36 Hz0.18 Hz0.09 Hz

0.026 Hz0.011 Hz

1.1 Hz

Ave

rage

wav

elet

tran

sfor

m

1.1 Hz0.11 Hz0.03 Hz

0.014 Hz

0.01 0.03 0.1 0.3 1

1.1 Hz0.18 Hz0.091 Hz0.032 Hz0.012 Hz

Frequency (Hz)

Respiration

ECG

HRV

BP

BF-1

BF-2

Note –

Log frequency resolution

Same spectral peaks inall data?

Peaks are broadened bytheir time-variations.

The body is “humming” withrhythms! But where do theycome from?



Conclusion


Physiological origins of the rhythms?

∼Hz Process1.0 Heart – obvious0.2 Respiration – obvious?0.1 Myogenic activity of smooth muscle – same in vitro0.04 Neurogenic activity – absent after denervation0.01 Endothelial vasodilation – NO-dependent0.007 Endothelial vasodilation – endothelin-dependent

Our interest centres especially on –

Relative amplitudes

Couplings

between respiratory, cardiac, and cortical (EEG) oscillations,which seem to reflect the state of the organism...



Conclusion


Anæsthesia – practice & problems

What is (general) Anæsthesia?

A chemical perturbation of the organism resulting in atemporary loss of consciousness.

Mysterious – mind/body relationship etc.

Certain nervous pathways blocked.

Important – needed for surgery.

“...the anæsthetist is still unable to measure the depthof anæsthesia in order to prevent inadvertent awakeningduring anæsthesia.”

C.J.D. Pomfrett, Brit. J. Anæsthesia, 1999.



Conclusion


Incidence of awareness in anæsthesia

Too much anæsthetic ⇒ bad for patient.Too little anæsthetic ⇒ awareness.

Author Date Sample Awareness %Hutchinson 1960 656 1.2Harris 1971 120 1.6McKenna 1973 200 1.5Wilson 1975 490 0.8Liu et al 1990 1,000 0.2Lennmarken & Sandin 2004 1,238 0.9R. Coll. of Anæsth. (NAP5) 2014 3,000,000 0.005

So need a reliable measure of depth of anæsthesia.Can physics (of coupled oscillators) help? Maybe...



Conclusion


Example – cardio-respiratory synchronisation in rats

Initially, studied rats.

a Administer anæsthetic;b Measure heart-rate

& respiration;c Anæsthesia lasts

about 100 min.

Synchronisation effectsare much stronger duringanæsthesia.

Starts wearing off aftertypically 40–50 min.

Depth of anæsthesia relatedto synchronisation ratio.

[Stefanovska et al., Phys. Rev. Lett. 85, 4831 (2000).]

0

0.5

1

λ 1,2

0

0.5

1

λ 1,3

0 10 20 30 400

0.5

1

λ 1,5

Time (min)

0

0.5

1λ 1,

4

2π

Ψ1

π

0



Conclusion


Aims and challenges

Interested in how to characterise signals so as to revealchanges in anæsthesia. Challenges include –

Numerous signals are potentially relevant – so which willbe most useful?

Inherent time-variations of the signals – how toaccommodate or exploit them?

How best to combine the information coming from manydifferent analyses?

New analysis methods needed to be developed beforeBRACCIA could come to fruition.



Conclusion


Data analysis methods

Wavelet analysis, the main “work horse”, provides –

Time-frequency information with logarithmic frequency resolutionPower in different spectral ranges

Frequency variability analysis – measured directly from ECG R-peaksand respiration signal maxima, giving HRV and RFV.

Wavelet phase coherence analysis – apply to any pair of signals. Ifphase coherence persists over time then –

Either the signals have a common sourceOr they are synchronised

Coupling function analysis – apply to any pair of signals. Extractsdetailed quantitative information about the mutual interactions betweenthe oscillators.

Result: a large number of different measures of the state, e.g. averages,powers in different spectral ranges for several signals, HRV, RFV, coherencesat different frequencies between different signals, coupling functions...



Conclusion


Making optimal use of the results – classification

Problem is how best to combine the diverse measures (attributes) ofstate, e.g. awake, or anaesthetised with propofol, or with sevoflurane.

Use distance-based classification and nearest neighbour classifier.

With an appropriate distance measure in the multi-dimensional space,the three states form separated clusters.

Leads to a confusion matrix giving the likelihood of correct and incorrectclassification for subjects in each group.

Using a learning algorithm, the classification accuracy improves withnumber of subjects.

[Kenwright et al, Anæsthesia 70 1356–1368 (2015), Suppl. Mat.]



Conclusion

ConsortiumResearch programme

Proposed BRACCIA Consortium

No. Organisation Abbreviation Town Country1 University of Ljubljana, UNILJFE Ljubljana Slovenia

Faculty of Elec. Eng.2 Lancaster University UNILANCS Lancaster UK3 Royal Lancaster Infirmary MBHT Lancaster UK4 Swiss Federal Inst, of Tech. EPFL Lausanne Switzerland5 University of Ljubljana, UNILJFM Ljubljana Slovenia

Faculty of Medicine6 University of Oslo UOUH Oslo Norway

Ulleval Hospital7 University of Potsdam UP Potsdam Germany

Inst. of Complex Systems8 Academy of Sciences ICSASCR Prague Czech Repub.

Inst. of Computer Sciences

Coordinator: Aneta Sefanovska



Conclusion


Measurements

The simultaneous measurements included time series of –

Electrical activity of the heart (ECG)

Respiratory activity

Skin temperature

Skin conductivity

Brain activity (EEG)

For patients (a) awake, (b) anæsthetised with propofol, (c)anæsthetised with sevoflurane.

All measurements used the Cardo&BrainSignals signalconditioning unit specially designed for BRACCIA by JozefStefan Institute (Ljubljana).



Conclusion


The Cardo&BrainSignals signal conditioning unit

Physiologicaltransducer

Signalconditioning

Anti-aliasingLP filter

DS ADCDecimationLP filter

Log fileF : 9600 HzOversamp. rat.: 128Resolution: 24-bitLP FIR filt. F =4.9 kHz

s

0

Fs: 1200 Hz

Main features –

12 identical channels

Synchronized parallel operation of ∆Σ ADCs

24-bit A/D conversion

9.6 kHz sampling frequency

Optical interface, to reduce interference

Same measurement set-up in both hospitals, as well as forhealthy subjects (Lancaster) and rats (Ljubljana).



Conclusion


Measurements

Channel Signal1 ECG2 EEG-13 EEG-24 EEG-35 EEG-46 BP Pulse7 Respiration8 Conductivity9 Temperature 1

10 Temperature 211 Gen. purpose 112 Gen. purpose 2

All recorded (i) without and(ii) with anæsthesia.

1

0

−1 EC

G [V

]

0

50

Res

p [r

ange

%]

0.52

0.54

Con

d [m

S]

0 0.3

Pul

se [V

]

30

35

Tem

p [° C

]

0

50 E

EG

1 [µ

V]

0

50

EE

G2

[µV

]

0

50

EE

G3

[µV

]

650 655 660 665 670 6750

50

EE

G4

[µV

]

Time [s]



Conclusion


BRACCIA Project Plan

Logistics

Measurements in Oslo,Lancaster, Ljubljana.

Resultant time seriesuploaded to database inLancaster.

Data downloaded toLancaster, Lausanne,Potsdam, Prague, foranalysis.

Joint, collaborativeinterpretation

Schedule

Proposal 2005-2008Extended 2005-2009Actual 2005-2016...

WP3:

Lead contractor:

Methodology: measurementsetups for simultaneous recordingof EEG, ECG, respiration, skintemperature and blood pressure

CO1, CR2

WP4:

Lead contractor:

Measurementson humans setupsfor simultaneousrecording of EEG,ECG, respiration,skin temperature andblood pressure

CO1, CR2, CR3,CR6

WP12:

Lead contractor:

Disseminationall groups

WP11:

Lead contractor:

Summary and planningthe protocol for a large-scalemulti-centre study

all groups

WP9:

Lead contractor:

Modeling brainwaves using a largeensemble ofoscillators

CO1, CR2, CR4, CR7

WP8:

Lead contractor:

Modeling cardiovascularinteractions using up to fivecoupled oscillators

CO1, CR2,CR7

WP7:

Lead contractor:

Neurological,physiological andclinicalinterpretation ofobtained results

all groups

WP6:

Lead contractor:

Analysis ofmeasured time-series

CO1, CR2, CR4,CR7, CR8

WP5:

Lead contractor:

Measurementson rats

CO1, CR5

WP

1:

Le

ad

co

ntr

ac

tor:

Pro

ject

ma

na

ge

me

nt

CO

1

WP2:

Lead contractor:

Methodology: developmentof methods for testing causalities(synchornization, inter-oscillatorscouplings, directionality,surrogates)

CO1, CR2,CR7, CR8

START

END

WP10:

Lead contractor:

Modelinginteractions on microand macroscopic level:between oscillators inthe brain, the cardiacand respiratoryoscillators

CO1,CR2, CR4, CR7

WP

1:

Le

ad

co

ntr

ac

tor:

Pro

ject

ma

na

ge

me

nt

CO

1



ConclusionSummary

Summary – not including outcomes

Physiological oscillations carry information about the stateof the organism.

Powerful time series data analysis methods are nowavailable to analyse the oscillations.

BRACCIA uses these to explore whether the oscillationscan quantify depth of anæsthesia.

Following about 11 years of work, the BRACCIA enterpriseis coming to fruition and a clear answer is now emerging...

...to be reported in the presentations by Andy Smith andJohan Ræder!



ConclusionSummary

Acknowledgements and recent publications

AcknowledgementsWe are grateful to the entire BRACCIA team for their collaboration, and toEuropean Community (FP6), the Engineering and Physical SciencesResearch Council (UK) and ARRS (Slovenia) for funding the research.

Recent BRACCIA publications1 D A Kenwright, Bernjak, T Draegni, S Dzeroski, M Entwistle, M Horvat,

P Kvandal, S A Landsverk, P V E McClintock, B Musizza, J Petrovcic, JRæder, L W Sheppard, A F Smith, T Stankovski and A Stefanovska,“The discriminatory value of cardiorespiratory interactions indistinguishing awake from anæsthetised states: a randomisedobservational study”, Anæsthesia 70 1356–1368 (2015).

2 T Stankovski, S Petkoski, J Ræder, A F Smith, P V E McClintock, and AStefanovska, “Alterations in the coupling functions between cortical andcardio-respiratory oscillations due to anæsthesia with propofol andsevoflurane”, Phil. Trans. R. Soc. A 374, 20150186 (2016).


Documents

Brain, Respiration, and Cardiac Causalities in Anæsthesia · 2016-04-15 · Brain, Respiration, and Cardiac Causalities in Anæsthesia Origins, basis, rationale and science Peter