b. more biopotentials - university of...

Biomedical Instrumentation B18/BME2

Biomedical

Instrumentation B. More Biopotentials

B18/BME2

Dr Gari Clifford

(Based on slides from

Prof. Lionel Tarassenko)


Diagnostic uses of ECG Foetal monitoring (both before birth & during)

Patient monitoring in Ambulance, Intensive Care

Unit or Coronary Care Unit

S-T segment elevation to diagnose heart attacks

Evidence of cardiac muscle damage (infarct)

Detection of precursors to heart attacks:

Abnormal heart beats (e.g. many ectopic beats, TWA)

Abnormal heart rhythms


Use of ECG in CCU

The ECG is highly informative in the diagnosis of a heart

attack (Myocardial Infarct). Insufficient blood supply to the

cardiac cells due to a blockage in the coronary arteries

(ischaemic heart condition) causes S-T segment elevation.

Following the heart attack, cardiac muscle damage (infarct)

generally leads to a loss of amplitude in the ECG.


ECG abnormalities

(possible precursors to heart attacks)

Analysis of the ECG can provide early warning of potential

problems.

Ectopic beats originate somewhere other than the Sino-

Atrial (SA) node and often have different shapes

(morphologies).

Abnormal heart rates (arrhythmias) can be treated.


Other intervals in ECG analysis

The most important interval in the ECG is the QT interval

A longer than normal QT interval is a good indicator of

long QT syndrome (LQTS)


Q-T interval measurement

LQTS is a potentially fatal condition that renders

sufferers vulnerable to an arrhythmia known as torsade

de pointes.

When this rhythm occurs the heart is unable to beat

effectively and the blood flow to the brain falls

dramatically.

The result is a sudden loss of consciousness and

possible cardiac death.


Detecting ECG abnormalities

Two methods are in common use:

Ambulatory monitoring

Exercise stress ECGs


Ambulatory ECG monitoring

ECG monitored for 24 hours.

Results printed out:

24-hour summary detailing the heart rate and S-T

segment changes over the period of the test.

Detailed information on ECG recorded at the time

of a significant event (e.g. arrhythmia).


Analysis of ECG waveform

Diagnostic information can be obtained by

analysis of the amplitude and relative timing

of the various segments.

The simplest interval to measure is the R-R

interval (from which the heart rate is derived).

Two types of heart rate meters:

Averaging heart rate meter

Beat-to-beat heart rate meter


Heart Rate Meters

Heart rate is usually given in beats per minute

(BPM).

The easiest way to obtain this is to count an

identifying feature in the ECG which occurs once

per heart beat.

The most obvious such feature is the QRS

complex which is a sharp spike.

Both averaging and beat-to-beat devices need

to perform this detection.


QRS detection

There are 4 main problems in detecting the QRS

complex in ECG traces:

Artefacts due to electrode motion


QRS detection

There are 4 main problems in detecting

the QRS complex in ECG traces:

Artefacts due to electrode motion

Baseline wander (mostly caused by breathing

and torso movements)

Muscle artefact (broadband)

T-waves with high-amplitude content


QRS detection

The solution to these problems is to use a band-

pass filter to remove:

Low-frequency changes such as baseline wander

High-frequency changes e.g. movement/muscle artefact

Most of the frequencies in the QRS complex are

around 5-20 Hz.

A pass-band of 10 – 40 Hz is therefore

appropriate.... Why?


Look at the signal below 10Hz .. Why do we want to remove it?


QRS detection

Once the “non-QRS” sections of the ECG

have been attenuated, the QRS complex

can be detected with a threshold detector.


R-Wave pulse generator

This should trigger a pulse generator so

that a short pulse of a fixed duration is

generated once (and only once) for each

QRS complex.



The “average power” of the pulse train from the

pulse generator circuit will be indicative of the

Heart Rate.

This can be determined using a “leaky integrator”

(a form of low-pass filter).

The time-constant of the R-C circuit should be

several beats long to minimise output ripple.



This is best achieved using a digital circuit which:

- Counts the time between consecutive QRS complexes

- Inverts this in order to obtain a heart rate (rather than interval)


Heart rate variability

Under resting conditions, the heart rate of a healthy individual is not

constant. (Notice compressions and rarefactions above)

During expiration, the vagus nerve is stimulated, which slows down

the heart rate (the right vagus innervates the sinoatrial node).

During inspiration, the vagus nerve is not stimulated.

This gives rise to a phenomenon known as respiratory sinus

arrhythmia (RSA); cardio-acceleration during inspiration, cardio-

deceleration during expiration.


Autonomic Regulation

Rest & Digest Fight & Flight


Heart rate variability

Upper trace: respiration rate from electrical impedance

plethysmography – see next lecture.

Middle trace: beat-to-beat R-R interval.

Lower trace: R-R interval series re-sampled at 4Hz and

cubic spline fitted to time series (smoothing).


The sympathovagal balance:

Ratio of LF power to HF power in PSD of heart rate

time series is though to reflect sympathovagal balance


The circadian rhythm


HRV and sleep state

HRV is also circadian

Autonomic balance

changes over 24 hours

Significantly in different

sleep cycles

Also changes based on

disease.

Wakefulness Deep Sleep

Light Sleep REM (Dream) Sleep

• HR & HRV are not specific enough to identify sleep stages

• So what is ...?


Other biopotentials

There are other biopotentials which can be

recorded from the body using similar circuitry:

The Electroencephalogram (EEG, electrical activity of

the brain)

The Electromyogram (EMG, electrical activity of

muscle)

The Electro-oculogram (EOG, electrical activity of the

eyes)

All are used in sleep staging


Brief intro to the Electroencephalogram

The EEG signal is also

measured with Ag-AgCl

electrodes

Placed in standard positions

on the scalp

Signal is <100μV – Why?

(Recall ECG is ~1mV)

Heart: ~3x109

Brain: ~1011

Due to skull attenuation


Characteristics of the EEG The important information is in the frequency domain.

The frequency range from 0.5 to 30 Hz has been

arbitrarily divided into 5 bands:

Delta 0.5-4Hz Deep Sleep

Theta 4-8 Hz Drowsiness / light sleep

Alpha 8-13 Hz Relaxed

Beta 13-22 Hz Alert

Gamma 22-30 Hz Short term memory tasks?


Diagnosis use of EEG

EEG helps the diagnosis of brain death, epilepsy and sleep

disorders

EEG during an epileptic seizure

10-20 Montage


Sleep analysis

Quality of life is heavily dependent on quality of sleep.

Between 5 and 10% of the adult population suffers from some form

of sleep disorder (insomnia, heavy snoring, Obstructive Sleep

Apnoea (OSA), etc…)

Such people may be referred to a “sleep clinic” by their GP where

various signals, including four channels of EEG, the EOG and

oxygen saturation, will be recorded throughout the night.

The EEG and the other signals are printed out and reviewed by a

trained sleep technician (requiring 2 to 5 hours for each record).


Sleep EEG The channels of sleep EEG are analysed using a rule-

based system

Consecutive 30s segments are assigned to one of six

stages (to form hypnogram)

Wake, Stage 1, Stage 2, Stage 3, Stage 4 & REM sleep

1= light, 3 & 4 = deep sleep

(Recently stages 3 &4 merged)


EEG rules for scoring ...

For example, two rules for stage 3:

an EEG record in which at least 20% but not

more than 50% of the epoch consists of waves of

frequency 2 Hz or lower which have amplitudes

greater than 75 μV peak to peak.

sleep spindles may or may not be present in

stage 3.


EEG in different sleep stages


Sleep structure is age-dependent

Over the night:

REM sleep

duration increases

& SWS decreases

Sleep changes in

adolescence

As we age this

pattern fragments


Automating sleep analysis

The important information is in the

frequency domain.

Use the Short-term Fourier Transform or

an Auto-Regressive (AR) model to extract

the frequency-domain information.


Automated sleep analysis Work performed in the 90’s and 00’s (in this Department)

has led to methods for analysing sleep on a 1s basis.

Sleep is treated as having three states:

Wakefulness, REM/light sleep, deep sleep

Sleep-wake continuum is represented by interpolation

between these states.


Short-term Fourier transform

First extract N samples of signal and then window

(using Hamming, Kaiser or Hanning windows) to

avoid sharp discontinuities at edges.

Then apply the Discrete Fourier Transform

[O(N2) operations] or the Fast Fourier Transform

[O(N log N) operations] if N is a power of 2.


AR models for spectral estimation

The notation AR(p) refers to the autoregressive model of

order p. The AR(p) model is written as follows:

X t = ai X t-i + t (1 i p)

where the ai’s are the parameters of the model and εt is a white-noise process with zero mean.

An autoregressive model is essentially an infinite

impulse response filter which shapes the white-noise

input. The poles are the resonances of the filter and

correspond to the spectral peaks in the signal.

http://en.wikipedia.org/wiki/Infinite_impulse_response

http://en.wikipedia.org/wiki/Infinite_impulse_response


AR-model vs FFT spectra (for EEG)

AR model is

parametric

Requires only a

few coefficients

Useful for

estimation on

short time series


Automated sleep analysis AR model parameters inputted to a neural network

Sleep is treated as having three states:

Wakefulness, REM/light sleep, deep sleep

1s epochs – continuous scoring ... But it maps to sleep

stages too ...


EMG

The electromyogram is used to identify

muscle activity

In sleep is it used to identify mastication

Eye flicks are not constant, so EMG under

chin increases confidence in REM score


EOG

The electro-oculogram is used to identify

rapid eye movement (indicates dreaming)

Try it in the lab!


Complex example - OSA

SNA: Sympathetic Nerve Activity (recorded from peroneal nerve)


EEG is also used for sedation

Look at coherence between different regions of

the brain

E.g. BIS monitor

Scale of 1:100 ... Proportional to hypnotic dose of

intravenous or volatile agents used, correlating well with

the hypnotic state and importantly is agent independent.

Does not identify movement or non-movement

response, especially in the presence of opiates

Anaesthesia is more than just a loss of

consciousness


Evoked potentials

This is a technique whereby a stimulus, such as a light

flash or loud click, is repeatedly applied.

The EEG signal is recorded from a particular area of the

brain.

Normal EEG activity, however, masks the brains response

to a single stimulus.

Repetitive stimuli have to be used and the evoked

response is distinguished from the background activity by

using the technique of signal averaging.

b. more biopotentials - university of...

Documents