Multimodal Pressure-Flow Analysis to Assess Dynamic Cerebral Autoregulation

1
Multimodal Pressure-Flow Analysis to Assess
Dynamic Cerebral Autoregulation
Albert C. Yang, MD, PhD Attending Physician,
Department of Psychiatry, Taipei Veterans
General Hospital, Taipei, Taiwan Assistant
Professor, School of Medicine, National
Yang-Ming University, Taipei, Taiwan
ccyang_at_physionet.org
2
Overview

• What is cerebral autoregulation and how to
  measure it?
• Multimodal pressure-flow analysis
• Empirical Mode Decomposition and Hilbert-Huang
  Transform
• Subsequent improvement
• Demonstration

3
Body as Servo-Mechansim Type Machine

• Importance of corrective mechanisms to keep
  variables in bounds.
• Healthy systems are self-regulated to reduce
  variability and maintain physiologic constancy.

Underlying notion of constant, steady-state,
conditions.
Walter Cannon 1929
4
Ideal Cerebral Autoregulation
Lassen NA. Physiol Rev. 195939183-238 Strandgaar
d S, Paulson OB. Stroke.198415413-416
5
Static Autoregulation Measurement
Tiecks FP et al., Stroke. 1995 26 1014-1019
6
Dynamic Autoregulation Measurement
Tiecks FP et al., Stroke. 1995 26 1014-1019
7
Autoregulation Index
Tiecks FP et al., Stroke. 1995 26 1014-1019
8
Challenges of Cerebral Autoregulation Assessment

• Blood pressure and cerebral blood flow velocity
  are often nonstationary and their interactions
  are nonlinear.
• Need a new method that can analyze nonlinear and
  nonstationary signals.

Novak V et al., Biomed Eng Online. 20043(1)39
9
Multimodal Pressure-Flow Analysis
10
Participants

• 15 normotensive healthy subjects
• age 40.2 2.0 years
• 20 hypertensive subjects
• age 49.9 2.0 years
• 15 minor stroke subjects
• 18.3 4.5 months after acute onset
• age 53.1 1.6 years

Novak V et al., Biomed Eng Online. 20043(1)39
11
Measurements

• Blood pressure
• Finger Photoplethysmographic Volume Clamp Method.
• Blood flow velocities (BFV) from bilateral middle
  cerebral arteries (MCA)
• Transcranial Doppler Ultrasound.

Novak V et al., Biomed Eng Online. 20043(1)39
12
Valsalva Maneuver
IV. increased cardiac output and increased
peripheral resistance
I. Expiration - mechanical
III. Inspiration - mechanical
II. reduced venous return, BP falls
13
Valsalva Maneuver Dynamics
Blood Pressure
Blood Flow Velocity Right Middle Cerebral Artery
Blood Flow Velocity Left Middle Cerebral Artery
14
Empirical Mode Decomposition (EMD)
? ? ?? Norden E. Huang

• The Empirical Mode Decomposition Method and the
  Hilbert Spectrum for Non-stationary Time Series
  Analysis, (1998) Proc. Roy. Soc. London, A454,
  903-995.
• The motivation of EMD development was to solve
  the problems of non-linearity and
  non-stationarity of the data
• Is an adaptive-based method

Cited 7,722 Times!
15
Empirical Mode Decomposition
Huang et al. Proc Roy Soc Lond A 1998454903-995.
16
Empirical Mode Decomposition
Step 1 Find the envelope alone local maximum and
minimum
Huang et al. Proc Roy Soc Lond A 1998454903-995.
17
Empirical Mode Decomposition
Step 2 Find the average between envelopes
Huang et al. Proc Roy Soc Lond A 1998454903-995.
18
Empirical Mode Decomposition
Step 3 To determine the fluctuation of original
signal around the average of envelopes
Huang et al. Proc Roy Soc Lond A 1998454903-995.
19
Empirical Mode Decomposition
Sifting to get all IMF components
Huang et al. Proc Roy Soc Lond A 1998454903-995.
20
Empirical Mode DecompositionA Simple Example
21
Empirical Mode Decomposition
Original blood pressure waveform
Key mode of blood pressure waveform during
Valsalva maneuver
22
Blood Pressure versus Blood Flow
VelocityTemporal (time) Relationship
Novak V et al., Biomed Eng Online. 20043(1)39
23
Blood Pressure versus Blood Flow VelocityPhase
Relationship
Control
Stroke
Novak V et al., Biomed Eng Online. 20043(1)39
24
Between Groups Phase Comparisons p lt 0.005,
p lt 0.01
Groups BPR Values Comparisons p lt0.001
25
(No Transcript)
26
(No Transcript)
27
Conventional Autoregulation Indices
Novak V et al., Biomed Eng Online. 20043(1)39
28
Summary Original Version of MMPF Analysis

• Regulation of BP-BFV dynamics is altered in both
  hemispheres in hypertension and stroke, rendering
  BFV dependent on BP.
• The MMPF method provides high time and frequency
  resolution.
• This method may be useful as a measure of
  cerebral autoregulation for short and
  nonstationary time series.

29
Limitations Original Version of MMPF Analysis

• Requires visual identification of key mode of
  physiologic time series
• Mode mixing with original EMD analysis
• Valsalva maneuver itself has certain risk

30
Subsequent Improvements of MMPF Analysis

• Use Ensemble EMD (EEMD) Analysis
• Resting-state MMPF Analysis
• Selection of key mode related to respiration
  during resting-state condition
• Comparison of phase shifts in multiple time
  scales
• Implementation and automation of the method

Wu, Z., et al. (2007) Proc. Natl. Acad. Sci.
USA., 104, 14889-14894
K. Hu, et al., (2008) Cardiovascular Engineering
M-T Lo, k Hu et al., (2008) EURASIP Journal on
Advances in Signal Processing
Hu K et al., (2012) PLoS Comput Biol 8(7)
e1002601
Dr. Yanhui Liu. DynaDx Corp. U.S.A.
31
Resting-State Multimodal Pressure-Flow Analysis
K. Hu, et al., Cardiovascular Engineering, 2008.
32
Respiratory Signals From Blood Pressure Time
Series
M-T Lo, k Hu et al., EURASIP Journal on Advances
in Signal Processing, 2008
33
Resting-State Multimodal Pressure-Flow Analysis
34
Resting-State Multimodal Pressure-Flow Analysis
35
Cerebral Blood Flow Regulation at Multiple Time
Scales
Hu K et al., PLoS Comput Biol 2012 8(7) e1002601
36
Traumatic Brain Injury and Cerebral Autoregulation
k. Hu, M-T Lo et al., journal of neurotrauma,
2009
37
Traumatic Brain Injury and Cerebral Autoregulation
k. Hu, M-T Lo et al., journal of neurotrauma,
2009
38
Midline Shift Correlates to Left-Right Difference
in Autoregulation
k. Hu, M-T Lo et al., journal of neurotrauma,
2009
39
(No Transcript)
40
Resources

• Empirical Mode Decomposition (Matlab)
• http//rcada.ncu.edu.tw/research1.htm
• DataDemon (Generic Analysis Platform)
• For 64-bit system,https//dl.dropbox.com/u/795530
  7/daily_build/x64/DataDemonSetupPRO.msi
• For 32-bit system,https//dl.dropbox.com/u/795530
  7/daily_build/x86/DataDemonSetupPRO.msi

41
Acknowledgements
Albert C. Yang, MD, PhD
Chung-Kang Peng, PhD
Vera Novak, MD, PhD
Ment-Zung Lo, PhD
Kun Hu, PhD
Yanhui Liu, PhD