ORIGINAL PAPER
什么东西助消化Design and Development of a Heart Rate Variability Analyzer
Aparna Mohan &Frana James &Sajeer Fazil &Paul K.Joph
Received:2June 2010/Accepted:13September 2010/Published online:6November 2010#Springer Science+Business Media,LLC 2010
Abstract Heart rate variability (HRV),analysis gives an insight into the state of the autonomic nervous system which modulates the cardiac activity.Here a digital signal controller bad handy device is developed which acquires the beat to beat time interval,process it using techniques bad on non-linear dynamics,fractal time ries analysis,and information theory.The technique employed,that can give reliable results by asssing heart beat signals fetched for a duration of a few minutes,is a huge advantage over the already existing methodologies of asssing cardiac health,tho being dependant on the tedious task of acquiring Electro Cardio Gram(ECG)signals,which in turn requires the subject to lie down at a stretch for a couple of hours.The nsor ud,relies on the technique of Photoplethysmography,rendering the whole approach as noninvasive.The device designed,calculates parameters like,Largest Lyapunov Exponent,Fractal dimension,Correlation Dimension,Approximate Ent
ropy and α-slope of Poincare plots,which bad on the range in which they fall,the cardiac health condition of the subject can be assd to even the extend of predicting upcoming disorders.The design of heart beat nsor,the technique ud in the acquisition of heart beat data,the relevant algorithm developed for the analysis purpo,are prented here.周公解梦梦见捡钱
Keywords Digital signal controller .Heart rate variability .Non-linear time ries analysis .Photoplethysmography
Introduction
Heart rate variability or HRV refers to the beat-to-beat fluctuations in heart beat,or more precily,to the variation in the intervals between the R points of the ECG,that is “RR intervals ”.HRV is a reliable reflection of many physiological factors modulating the normal rhythm of the heart [1].Its variation may contain indicators of current dia,or warnings about impending cardiac dias.HRV is a non stationary signal and is found to exhibit the property of deterministic chaos and hence techniques bad on non-linear dynamics,fractal time ries analysis,and information theory are ud here to analyze it.
In the first pha,a low cost,low power,non-invasive heart beat nsor is developed using the techniq
ue of photoplethysmography.The filtered,amplified and digi-tized signal is fed into a digital signal controller,which extracts the R-R time interval and displays it in a 16×2alphanumeric LCD screen.After enough amounts of data (700data values, 1.4kb)is stored in Random Access memory (RAM),it is analyzed using the algorithm developed.The parameters,Lyapunov exponent,Fractal dimension,Correlation dimension,alpha value from Poin-care plot,slope of Detrended Fluctuation Analysis and Approximate Entropy are calculated and displayed.The range into which the parameters fall can be ud for asssing the cardiac health of the subject.
The technological advancements happening in the area of embedded systems,give us a more practical and effective way of implementing real time systems,especially in the medical field.In this paper,the advantages of using a Digital Signal Controller,produced by Microchip,are utilized in the best possible way,to replace the conventional platforms like MATLAB,Mathematica etc.Thus,the
A.Mohan :F.James :S.Fazil :P.K.Joph (*)
Department of Electrical Engineering,National Institute of Technology Calicut,Calicut,India
e-mail:paul@nitc.ac.in
J Med Syst (2012)36:1365–1371DOI 10.1007/s10916-010-9597-6相关英语
system gets converted to a handy device,which carries more advantages,and it can be even improved in future,for predicting the health condition of a patient.
Heart beat nsor
Heart rate variability
Heart Rate Variability refers to the variation in the intervals between the R points of the Electro Cardio Graph,the“RR intervals”.The clinical relevance of HRV has been established through a large number of experiments that point out to the ability of HRV as a quantitative marker of autonomic nerve activity,which in turn has a very clo association with cardiovascular activity[1–4].The propen-sity for lethal arrhythmias is believed to be cloly related to signs of either incread sympathetic or reduced vagal activity.Numerous papers are available in literature that shows reduced HRV associated with various cardiac and non-cardiac dias like,myocardial infarction,Diabetic Neuropathy,myocardial dysfunction,tetraplegia,etc.[5–9].
Photoplethysmography
Photoplethysmography(PPG)involves measurement of blood volume pulsations in peripheral arteries
and capillaries by detection and temporal analysis of tissue-scattered optical radiation[10].Arterial blood volume changes according to heart beat and this leads to change in optical properties of the tissue,which can be measured by PPG.A PPG is often obtained by passing infra red radiations which illuminates the skin and changes in light absorption can be measured using a nsor,which can be converted into current. Acquisition of PPG
The major components in PPG system are,photo source, photo detector,signal amplifier and signal filters.An infrared Light Emitting Diode(LED)illuminates the tissue and a light nsitive detector(LSD),which has been tuned to the same color frequency as the LED,detects the amount of light transmitted through the tissue.The LED and the LSD are mounted in a spring-loaded device that can be clipped onto the fingertip or ear lobe.The OPT101,a monolithic photodiode with on-chip trans-impedance am-plifier,is ud in this work for detecting pul rate variations.The integrated combination of photodiode and trans-impedance amplifier on a single chip eliminates the problems commonly encountered in discrete designs such as leakage current errors,noi pick-up and gain peaking due to stray capacitance.Instrumentation amplifier
As the next step,an instrumentation amplifier is ud to extract and amplify the low level ECG signal(0–10mV) from a relatively high level(-hundreds of mV)of common mode interference,electrod
e artifact etc.The design considerations are,
&Constant gain:5–100
&Common mode rejection ratio(CMRR):100dB The design is shown in Fig.1.
Filtering
The main noi components prent in the output,such as baline wander,power line interference and muscle noi are filtered using high pass filtering,notch filtering and low pass filtering respectively.The design considerations are, &Sallen key low pass filter(shown in Fig.2):
Cutoff frequency:80Hz
Constant gain:1.6
&Sallen key high pass filter(shown in Fig.3): Cut off frequency:.5Hz
Constant gain:1.6
&Notch filter(shown in Fig.4):
Noi removal at50Hz
Digitization
The output of the filter is pasd to a comparator to detect peaks.The comparator output becomes high every time
a
Fig.1Instrumentation amplifier
长字成语
番茄营养
peak occurs and remains low otherwi.An LM311comparator is ud.The square wave generated as the output of the comparator is given to one of the interrupt pins of a digital signal controller.
The digitized output waveform,which is fed into the Digital signal controller (DSC),is shown in Fig.5.
Algorithm for finding RR interval
The digitized output is fed into the Capture/Compare/Pul (CCP1)pin of dsPIC33FJ128MC802,a digital signal controller manufactured by Microchip.The pul acts as an external interrupt on its falling edge.In the interrupt rvice routine,a timer has been initialized and it starts counting until it overflows.On overflow,the program goes to the timer interrupt rvice routine,where a counter variable is incremented each time the timer overflows.The time duration between two successive interrupts is calcu-lated in the program,which is stored in the RAM and simultaneously displayed in the Liquid crystal display (LCD).The data values are displayed in milliconds with four digits of accuracy.
Analysis of HR data
Nonlinear deterministic dynamic approach,bad on the theory of chaos,provides a description about many complex phenomenons in the real world,showing that apparently erratic behavior can be generated even by simple deterministic system with non linear structure [11].More-over,the intrinsic unpredictability can be explained by their strong dependence on initial conditions.
Nonlinear analysis of time ries provides a parameter t that quantifies the characteristics of the system attractor even when we do not know the model structure.This is particularly uful in the study of heart dynamics,as it has to start from external obrvation of the system.The various techniques ud in nonlinear dynamics are dealt with in the subquent ctions.Pha space reconstruction
Given with a time ries,an interrogation into the original pha space characteristics of the system generating it is done by the technique of pha space reconstruction.The embedding theorem propod by Taken ’s assures that the topological properties of the original system is prerved onto the reconstructed pha space [12,13].For a scalar time ries R −R (t),where t
=
Fig.5The digitized output signal of the
nsor
1,2…N,the time delay vectors in pha space can be reconstructed as defined:
X t¼RR tðÞ;RR tþt
ðÞ;RR tþ2t
ðÞ;:::;RR tþmÀ1
ðÞt
ðÞ½
ð1ÞWhere,τis referred to as the delay time and m is the embedding dimension.The dimension m of the recon-structed pha space is considered as the sufficient dimension for recovering the object without distorting any of its topological properties.
Largest Lyapunov exponent
Lyapunov exponents quantify the average exponential rate of paration between pha space trajectories with nearby initial conditions,nsitivity to initial conditions being the soul of chaos.
The spectrum of Lyapunov exponents,λi(i=1,2,...,n),λi s arranged in the descending order,is obtained by considering a small n-dimensional sphere of initial con-ditions,where n is the number of equations(or,equiva-lently,the number of rate variables)ud to describe the system.As time(t)progress,the sphere evolves into an ellipsoid who principal axes expand(or contract)at rates given by the Lyapunov exponents.The i th Lyapunov exponent,λi is defined as,
i = lim log
λ [p i (t))/ (p i (0)]
t→∞
ð2Þ
Where,p i is the i th principal axis.Λ1is the largest lyapunov exponent(LLE),which is estimated here using the algorithm propod by Wolf et al[14].A positive value for LLE for all initial conditions is a sure certificate for proving the deterministic chaos nature of the system. Fractal dimension(FD)
The pha space reprentation of a nonlinear,autono-mous,dissipative system can contain one or more attractors with generally fractional dimension.This attractor dimension is invariant,even under different initial conditions,which explains why the FD of attractors has been ud widely for system characteriza-tion.However,estimating the FD of the attractors involves a large computational burden.An embedding system has to be constructed from the original time-domain signal,bad on the method of delays and the attractor of this system has to be untangled before estimating its FD[15].Using box counting technique,the algorithms propod are higuchi’s algorithm,katz’s algorithm,Petrosian’s algorithm of which the former is ud here.
Correlation dimension(CD)
彩虹怎么画
Compared to other methods of measuring dimension,the correlation dimension,which is bad on the Grassberger–Procassia Algorithm,has the advantage of being straight-forward and quickly calculated,and is often in agreement with other calculations of dimension[15,16].The CD value will be high for chaotic data and it decreas as variation of RR signal becomes less and rhythmic.
掠夺造句
Approximate entropy(ApEn)
First propod by Pincus in1991,approximate entropy is a statistic that can be ud as a measure to
quantify the complexity of a signal[17].Its significance in HRV analysis is due to its properties of high resistance to short strong transient interference,suppression of noi and giving robust estimate for shorter data-in the range of 100–5,000points[17,18].
Poincare plots
党的优良作风The Poincaréplot is a tool developed by Henri Poincaréfor analyzing complex systems.In the context of medical sciences it is mainly ud for quantifying the heart rate variability(HRV)and proves to be quite an effective measure of this marker.The plot provides summary information as well as detailed beat-to-beat information on the behavior of the heart[19,20].The plot consists of points formed by each pair of successive RR interval data. The technique of ellip fitting is employed for character-izing the plot into functional class,whereby the ratio of
Fig.6Block diagram reprentation of the system developed
minor axis to major axis is computed.This value is an indicator of degree of heart failure in a subject.This ratio shows the ratio of short interval variation to the long interval variation.This ratio is more in the ca PVC (preventricular contraction),AF (arterial fibrillation),SSS (Sick Sinus Syndrome)due to high RR variations.But,this ratio falls (below normal)for the slowly varying signals like CHB (Complete Heart Block),Ischemic/dilated car-diomyopathy [21,22].
Detrended fluctuation analysis (DFA)
Detrended fluctuation analysis,is a method for determining the statistical lf affinity of a signal,in stochastic process,chaos theory and time ries analysis.The method of DFA has proven uful in revealing the extent of long-range correlations in time ries.The DFA can be ud to quantify the fractal scaling properties of short interval RR interval signals.This technique is a modifica-tion of root mean square analysis of random walks applied to non stationary signals [23].Fluctuations are character-
ized by a scaling exponent (lf similarity factor),αthat is the slope of the linear regression between fluctuation and window size.
In this method,a fractal like signal results in a scaling exponent value of 1,while Gaussian noi (totally random signal)results in a value of 0.5,and a Brownian noi signal with spectrum rapidly decreasing power in the higher frequencies results in an exponent value of 1.5[24,25].The can be viewed as an indicator of the “roughness ”of the original time ries;the larger the value of the α,the smoother the time ries.This slope is very low for very highly varying signals like PVC,AF and VF etc.But for rhythmically varying signals like SSS,CHB and ischemic/dilated cardiomyopathy,this value is slightly higher [21,22].
Implementation Hardware
The model of the system developed is shown in Fig.6
.
Fig.7Circuit designed in Proteus for simulation
Table 1Parameters values obtained for different health conditions Health condition LLE FD CD ApEn Alpha of Poincare
Slope of DFA Normal 0.64±0.09 1.34±0.17 1.67±0.19 1.21±0.080.5±0.120.85±.14CAST
0.25±0.07.97±0.11 1.21±0.15 1.01±0.070.2±0.08 1.08±0.09Congestive heart failure
0.21±0.09
.93±0.09
1.19±0.13
0.94±0.06
0.19±0.11
1.14±0.15
CAST cardiac arrhythmia suppression trial
