electrocardiogramm derived respiratory signal (edr signal)
TRANSCRIPT
1
EDST - UL Ecole Doctorale des Sciences
et de Technologie Université Libanaise
“Derivation of Respiratory Signal from
Electrocardiograms”
Submitted to Dr. Hassan Amoud & Dr. Ahmad Diab February 19th, 2016
Sarah Hussein Master TIS | TIS01 Course
O U T L I N E S
Context Problem Statement
Definition Signal Acquisition
Pre-Treatment
State of The Art Detailed Method
Brief Summary
INTRODUCTION EDR SIGNAL METHODOLOGIES CONCLUSIONS
1� 2� 3� 4�
2 Sarah Hussein Master TIS | TIS01 Course
INTRODUCTION
¨ Context ¨ Problem Statement 1�
3 Sarah Hussein Master TIS | TIS01 Course
I N T R O D U C T I O NContext
q Respiration and heartbeat are physiological functions critical for life
q Both are modulated by fluctuations of the autonomic nervous system (ANS)
q Both carry information for the autonomic control of the cardio-respiratory system
q Studies have shown a significant relationship between the ANS and
cardiovascular mortality, including sudden cardiac death
q Heart rate variability (HRV) is a simple and non-invasive method to asses the
status of the autonomic modulation of the heart
4
Sarah Hussein Master TIS | TIS01 Course
I N T R O D U C T I O NContext (Cont.)
q HRV is the amount of variations of both RR intervals and instantaneous heart rate
q A major component of HRV is respiratory sinus arrhythmia (RSA)
q RSA is the naturally occurring beat-to-beat fluctuation in heart rate that occurs
during a respiratory cycle
q Respiratory rate is crucial for the correct interpretation of RSA and thus HRV
q Respiratory signal is recorded with spirometry, pneumography or
plethysmography
5 Sarah Hussein Master TIS | TIS01 Course
I N T R O D U C T I O NContext (Cont.)
q These techniques require the use of cumbersome devices that may interfere with
natural breathing, and are unmanageable in certain applications
q The joint study of the respiratory and cardiac systems is of great interest
q The use of methods for indirect extraction of respiratory information is
particularly attractive to pursue
q It is more convenient to use physiological signal that does not alter respiration,
easily accessible & carries unambiguous information of respiration
q ECG is one such signal!
6 Sarah Hussein Master TIS | TIS01 Course
I N T R O D U C T I O NProblem Statement
“How can spontaneous respiration be derived from the Electrocardiograms?“
7 Sarah Hussein Master TIS | TIS01 Course
EDR SIGNAL
¨ Definition ¨ Signal Acquisition ¨ Pre-Treatment
2�
8 Sarah Hussein Master TIS | TIS01 Course
E D R S I G N A LDefinition | Respiratory Signal
q Respiration is an involuntary act controlled the ANS via the medulla oblongata of
the brain
q Medulla oblongata senses blood levels of CO2 and triggers respiration at
increased CO2 levels
q Respiratory system’s function is to meet the respiratory demands of cells:
¨ To deliver a steady supply of oxygen to the cells of the body
¨ To remove the carbon dioxide released by the cells
q During respiration, air is moved into & out of by changing the volume of lungs
q Lungs volume changes due to contractions of intercostal muscles & diaphragm
9
Sarah Hussein Master TIS | TIS01 Course
E D R S I G N A LDefinition | Respiratory Signal (Cont.)
10
F.H. Martini. Fundamentals of Anatomy and Physiology. Number ISBN: 0- 321-31198-1. Pearson, 7th edition edition, 2006
Sarah Hussein Master TIS | TIS01 Course
E D R S I G N A LDefinition | Respiratory Signal (Cont.)
11
q Autoregressive power spectra of respiratory signal (RESP) during controlled respiration at 6 (CR 6), 10 (CR 10) and 16 (CR 16) breaths/min in one subject
Maria Vittoria Pitzalis et al. Effect of respiratory rate on the relationships between RR interval and systolic blood pressure
Fluctuations: a frequency-dependent phenomenon Oxford Journals
Sarah Hussein Master TIS | TIS01 Course
E D R S I G N A LDefinition | EDR Signal
q EDR signals are the Electrocardiogram Derived Respiratory signals
q EDR methods exploit the respiratory induced changes of the ECG to provide a
surrogate respiratory signal
q These surrogate signals enable the estimation of the respiratory rate and the
temporal pattern of respiration
12
Komorowski et al. BioMed Eng OnLine (2015) 14:60
Sarah Hussein Master TIS | TIS01 Course
E D R S I G N A LSignal Acquisition
q A c q u i s i t i o n m e t h o d s o f E D R i s
consequently that of ECG
q ECG is measured via electrodes on the
surface of the body
q When the heart beats, a wave of
depolarization travels through the heart
q A potential difference is measured
between 2 points
13
F.H. Martini. Fundamentals of Anatomy and Physiology. Number ISBN: 0- 321-31198-1.
Pearson, 7th edition edition, 2006.
Sarah Hussein Master TIS | TIS01 Course
01 02 03
E D R S I G N A LSignal Acquisition (Cont.)
Limb Leads Augmented Limb Leads Precordial Leads
Three Lead Systems
Nobelprize.org. The electrocardiogram looking at the heart of electricity.
D. C. Randall and Y. M. El-Wazir. ECG Interpretation. Number ISBN 1593771800. Hayes Barton Press, 2004.
14 Sarah Hussein Master TIS | TIS01 Course
E D R S I G N A LPre –Treatment | System Requirements
q The detected ECG signal is noisy
q EDR extraction is done using ECG pure signal
q A pre-treatment of the signal is required to remove artifacts
q Pre-treatment should be performed on data that satisfies the following: ¨ The ECG should be recorded on equipment that satisfy voluntary standards ¨ The sample rate should be ≥ 500Hz
¨ The recordings of should be of 5 minute duration ¨ The environment of the recording should be controlled and physiologically stable
15 Sarah Hussein Master TIS | TIS01 Course
E D R S I G N A LPre –Treatment | Filtering & QRS Detection
q Pre-treatment includes :
¨ Filtering the ECG
¨ Detection of the QRS complexes in 2 stages
Lasse Sohrt-Petersen Evaluation of Algorithms for ECG Derived Respiration
in the Context of Heart Rate Variability Studies Aalborg University, 2014
16 Sarah Hussein Master TIS | TIS01 Course
E D R S I G N A LPre- Treatment | Filtered ECG Signal
F.H. Martini. Fundamentals of Anatomy and Physiology. Number ISBN: 0- 321-31198-1.
Pearson, 7th edition edition, 2006.
17 Sarah Hussein Master TIS | TIS01 Course
E D R S I G N A LPre-Treatment | Filtered ECG Signal
ECG Statistics, Noise, Artifacts, and Missing Data Gari D. Clifford
18 Sarah Hussein Master TIS | TIS01 Course
METHODOLOGIES
¨ State of The Art ¨ Detailed Method 3�
19 Sarah Hussein Master TIS | TIS01 Course
M E T H O D O L O G I E SState of The Art | EDR Algorithms
q Several EDR Algorithms were discussed in Literature:
¨ Multiple lead methods based on variations in angle of mean electrical axis (Moody
1985, Zhao 1994 & Caggiano 1996)
¨ Single lead methods based on R-wave amplitude or QRS area (Khaled 1992,
Felblinger 1997, Dobrev 1998, Mason 2001, de Chazal 2003, O’Brien 2007, Park
2008 & Arunachalam 2009)
¨ Methods based on heart rate (Womack 1971 & Correra 2008)
¨ Methods based on discrete Wavelet Transform and band pass filtering (Yi 2002 &
Boyle 2009)
¨ VCG methods based on variations in angle of mean electrical axis (Leanderson 2003
& Bailon 2003)
¨ Combinative methods (Varanini 1990, Orphanidou 2009 & Boyle 2009)
20 Sarah Hussein Master TIS | TIS01 Course
M E T H O D O L O G I E SState of The Art | EDR Algorithms (Cont.)
q Several EDR Algorithms were discussed in Literature:
¨ Multiple lead methods based on variations in angle of mean electrical axis (Moody
1985, Zhao 1994 & Caggiano 1996)
¨ Single lead methods based on R-wave amplitude or QRS area (Khaled
1992, Felblinger 1997, Dobrev 1998, Mason 2001, de Chazal 2003, O’Brien 2007,
Park 2008 & Arunachalam 2009)
¨ Methods based on heart rate (Womack 1971 & Correra 2008)
¨ Methods based on discrete Wavelet Transform and band pass filtering (Yi 2002 &
Boyle 2009)
¨ VCG methods based on variations in angle of mean electrical axis (Leanderson 2003
& Bailon 2003)
¨ Combinative methods (Varanini 1990, Orphanidou 2009 & Boyle 2009)
21 Sarah Hussein Master TIS | TIS01 Course
M E T H O D O L O G I E SState of The Art | Detailed Method
Single Lead Method Based on QRS Area
q EDR signal is obtained by calculating the area enclosed by the baseline and the
QRS complex in a certain interval
q The leads most affected by the presence of air in the lungs are Lead II and Lead V4
Why?
¨ The amplitude of the QRS complex depends on the ECG lead
¨ The leads most affected by normally air filled lungs are lead II & lead V4
q The QRS area approach is less affected by noise compared to pure amplitude EDR
methods
22 Sarah Hussein Master TIS | TIS01 Course
M E T H O D O L O G I E SState of The Art | Detailed Method (Cont.)
Power spectrum q Respiratory signal (solid
line) q EDR signals before
substitution of noisy beats (dotted line)
q EDR signals after
substitution of noisy beats (dashed line)
23
Raquel Bailon, Leif Sornmo, and Pablo Laguna. Chapter 8 – ecg derived respiratory frequency estimation.
Sarah Hussein Master TIS | TIS01 Course
M E T H O D O L O G I E SState of The Art | Detailed Method (Cont.)
Single Lead Method Based on QRS Area (Cont.) q Performance measures:
¨ Mean Square Error
¨ Correlation Coefficient : Pearson’s r
24 Sarah Hussein Master TIS | TIS01 Course
M E T H O D O L O G I E SState of The Art | Detailed Method (Cont.)
Single Lead Method Based on QRS Area (Cont.)
Lasse Sohrt-Petersen Evaluation of Algorithms for ECG Derived Respiration in the Context of Heart Rate Variability Studies Aalborg University, 2014
(RMSE=0.234 s & r=0.938)
25 Sarah Hussein Master TIS | TIS01 Course
CONCLUSIONS
¨ Brief Summary 4�
26 Sarah Hussein Master TIS | TIS01 Course
C O N C L U S I O N SBrief Summary
q Respiratory rate is crucial for the correct interpretation of heart diseases
q Respiratory signal is usually recorded with cumbersome devices
q Physiological signals such as ECG are preferable à EDR signals
q Signal Acquisition is done using 12 lead system
q Detected noisy signal requires filtering and QRS detection in two stages
q Single lead method based on QRS area detection exhibits good performance
27 Sarah Hussein Master TIS | TIS01 Course
THANK YOU
28
R E F E R E N C E S
q Sung-Bin park et al. An improved algorithm for respiration singal extraction from
electrocardiogram measured by conductive tectile electrodes using instantaneous frequency
estimation. Medical and Biological Engineering and Computing, 46:147–158, 2008
q Raquel Bail on, Leif So rnmo, and Pablo Laguna. Chapter 8 - ecg-derived respiratory
frequency estimation. In Gari D. Clifford, Francisco Azuaje, and Patrick E. McSharry, editors,
Advanced Methods and Tools for ECG Data Ana- lysis. Artech House, Inc, 2006. ISBN -10:
1-58053-966-1
q David Caggiano and Stanley Reisman. Respiration derived from the electro- cardiogram: A
quantitative comparison of three different methods. Proceedings of IEEE 22nd Annual NE
Bioengineering Conference, pages 103–104, 1996
29 Sarah Hussein Master TIS | TIS01 Course
R E F E R E N C E S
q Philip de Chazal et al. Automated processing of the single-lead electrocardiogram for
the detection of obstructive sleep apnoea. IEEE Transactions on Biomedical
Engineering, 50(6):686–696, June 2003
q A. Despopulos and S. Silbernagl. Color Atlas of Physiology. Number ISBN
3-13-545005-8. Thieme, 5th edition edition, 1998
q George B. Moody et al. Derivation of respiratory signals from multilead ecgs. Computers
in Cardiology, 12:113–116, 1985
q I. Kestin. Control of heart rate. Update in Anaesthesia, (3):15–16, 1993
30 Sarah Hussein Master TIS | TIS01 Course
Back Up Slides
31
E D R S I G N A LDefinition | Respiratory Signal (Cont.)
q The respiratory system is able to adapt to meet different levels of metabolic
need by increasing or decreasing the respiratory rate and volume
q Respiratory rate is the number of breaths taken within a set amount of time,
usually one minute
q Normal respiratory rate, called eupnea, ranges from 12 to 20 breaths per minute
in resting adults
q The respiratory volumes and capacities are the amount of air inspired, expired
and stored within the lungs
32 Sarah Hussein Master TIS | TIS01 Course
E D R S I G N A LAcquisition Methods (Cont.)
q Limb leads à Einthoven’s lead system à
Kirchoff’s voltage law (VI+VIII=VII)
q Augmented Leads à Wilson’s Central
Terminal à WCT=1/3(RA +LA+LL)
q Limb + Augmented leads = hexa-axial
reference system à calculate the electrical
axis of the heart in the frontal plane
Nobelprize.org. The electrocardiogram
looking at the heart of electricity.
33 Sarah Hussein Master TIS | TIS01 Course
E D R S I G N A LAcquisition Methods (Cont.)
q 6 Precordial leads à calculate the
electrical axis of the heart in the horizontal
plane
q These 6 electrodes as positive references
and WCT as negative reference =
precordial leads (V1, V2, V3, V4, V5 and
V6)
q Limb + Augmented leads + precordial
leads = 12 lead system à most commonly
used clinical ECG-system
D. C. Randall and Y. M. El-Wazir. ECG Interpretation. Number ISBN 1593771800. Hayes Barton Press, 2004.
34 Sarah Hussein Master TIS | TIS01 Course
E D R S I G N A LPre –Treatment | Pre – Filtering (1)
q The composite amplitude, R(t), of a detected R-wave can be modeled as
¨ r is the true R-wave amplitude during the resting phase of a normal breath ¨ a is the amplitude modulation due to respiration ¨ Common sources of noise include power-line interference (np), zero mean
Gaussian noise (nG), EMG noise (nH) and baseline wander (b)
q Due to the relation between baseline wander and respiration, the baseline wander
should not be much attenuated à HPF cut off frequency of 0.05Hz
q Assuming lowest breath frequency of 3 breaths per minute à normal breathing in
resting adults is 12 to 20 breaths per minute (0.2Hz to 0.33Hz)
35 Sarah Hussein Master TIS | TIS01 Course
E D R S I G N A LPre –Treatment | Pre – Filtering (2)
q np is caused by improper grounding of the ECG recording equipment or interference
from surrounding electronic equipment à 50Hz 2nd order IIR notch filter
Lasse Sohrt-Petersen
Evaluation of Algorithms for ECG Derived Respiration in the Context of
Heart Rate Variability Studies
Aalborg University, 2014
36 Sarah Hussein Master TIS | TIS01 Course
E D R S I G N A LPre –Treatment | Pre – Filtering (3)
q np is caused by improper grounding of the ECG recording equipment or interference
from surrounding electronic equipment à 50Hz 2nd order IIR notch filter
Lasse Sohrt-Petersen Evaluation of
Algorithms for ECG Derived Respiration in the Context of
Heart Rate Variability Studies
Aalborg University, 2014
37 Sarah Hussein Master TIS | TIS01 Course
E D R S I G N A LPre –Treatment | Pre – Filtering (4)
q EMG noise, nH (t), and zero mean Gaussian white noise, nG(t) à 2nd order Butterworth low-pass filter with a cut-off frequency of 40 Hz
Lasse Sohrt-Petersen
Evaluation of Algorithms for ECG Derived Respiration in the Context of
Heart Rate Variability Studies
Aalborg University, 2014
38 Sarah Hussein Master TIS | TIS01 Course
E D R S I G N A LPre –Treatment | QRS Detection
q Several QRS detection algorithms have been presented in the literature
q The Hamilton-Tompkins QRS detector is one of the most robust and proven QRS detector algorithms
q The original QRS Hamilton-Tompkins detector consist of a two stages:
¨ Stage 1: A preprocessing stage including filtering, differentiating, squaring, and time averaging
¨ Stage 2: A series of heuristic decisions rules which operates on the output of stage 1, in order to locate the QRS complexes in the original filtered ECG data
39 Sarah Hussein Master TIS | TIS01 Course
E D R S I G N A LPre –Treatment | QRS Detection | Phase 1
Lasse Sohrt-Petersen Evaluation of Algorithms for ECG Derived Respiration in the Context of Heart Rate Variability Studies Aalborg University, 2014
40 Sarah Hussein Master TIS | TIS01 Course
E D R S I G N A LPre –Treatment | QRS Detection | Phase 2
Lasse Sohrt-Petersen Evaluation of Algorithms for ECG Derived Respiration
in the Context of Heart Rate Variability Studies Aalborg University, 2014
41 Sarah Hussein Master TIS | TIS01 Course
M E T H O D O L O G I E SState of The Art | Respiratory-Induced Modulations of ECG
q To derive a surrogate respiratory signal from the information contained in ECG à Respiratory mechanisms that induce modulations of the ECG
q Respiratory activity influences the measurements of ECG in various ways:
¨ Respiratory induced modulation of the heart rate (RSA) leading to a frequency modulation of the ECG
Lasse Sohrt-
Petersen Evaluation of
Algorithms for ECG Derived Respiration in the Context of
Heart Rate Variability Studies
Aalborg University, 2014
42 Sarah Hussein Master TIS | TIS01 Course
M E T H O D O L O G I E SState of The Art | Respiratory-Induced Modulations of ECG (Cont.)
¨ The filling and empting of air in the lungs leads to changes in the transthoracic impedance which lead to an amplitude modulation of the ECG
¨ The mean electrical axis of the cardiac vector changes its direction during respiration, leading to both an amplitude and a frequency modulation of the ECG
Lasse Sohrt-Petersen Evaluation of Algorithms for ECG Derived Respiration
in the Context of Heart Rate Variability Studies Aalborg University, 2014
43 Sarah Hussein Master TIS | TIS01 Course