bme2020 2 ecg - umm.uni-heidelberg.de&kulvwldq 7|qqhv , 6olgh , *doydqlf lqvxodwlrq 1r hohfwulf...

10/30/2020

1

Biomedical EngineeringBioelectric Signals: EEG & ECG

Christian Tönnes

Christian Tönnes I Slide 2 I 30.10.2020

Schedule

Day Date Time Lecturer TopicTuesday 03.11.2020 13:00-14:30 Licht Biosensors & Physiological Signals ITuesday 10.11.2020 13:00-14:30 Licht Biosensors & Physiological Signals IIThursday 12.11.2020 13:00-14:30 Tönnes Bioelectrical Signals: EEG & ECGTuesday 17.11.2020 13:00-14:30 Reichert Medical Imaging: MRIThursday 19.11.2020 13:00-14:30 Tönnes Medical Imaging: CT, X-Ray & USTuesday 24.11.2020 13:00-14:30 Golla Medical Imaging: OtherThursday 26.11.2020 13:00-14:30 Andoh Blood Flow & Pressure ITuesday 01.12.2020 13:00-14:30 Andoh Blood Flow & Pressure IIThursday 03.12.2020 13:00-14:30 Golla Machine Learning ITuesday 08.12.2020 13:00-14:30 Golla Machine Learning IIThursday 10.12.2020 13:00-14:30 Golla 3D PrintingTuesday 22.12.2020 13:00-14:30 All Recap - Exam Preparations

Tuesday 26.01.2021 10:00-12:00Golla, Tönnes

Exam

Wednesday 03.02.2021 08:30-10:00Golla, Tönnes

Repeat Exam

10/30/2020

2


Materials

Slides are available from our website:https://www.umm.uni-heidelberg.de/inst/cbtm/ckm/lehre/


Lecturer: Christian Tönnes

PhD Student:

Medical Image Analysis with Deep Learning

Computer Assisted Clinical Medicine

Medical Faculty Mannheim

Heidelberg University

[email protected]

Phone:+49 (0) 621 383 4603

10/30/2020

3


MEASURING ECG/EEG/EMG


Circuit to measure electric signals

1. Electrode

2. Surge protection / galvanic insulation

3. Instrumentation Amplifier

4. Noise Filtering

5. Analog-to-digital converter

Right leg driver to reduce noise

http://openeeg.sourceforge.net

10/30/2020

4


Electrode

Connection between Circuit and Skin

Conductive gel for better signal

Fastened with glue or suction

Artifacts generated by:

• Movement of Electrode

• Bad conductivity

Glue Electrodes © Christian Tönnes

Suction Electrodeshttps://www.cardiodepot.eu/s/17769_158780_suction-electrode-system-dt-80-straessle


Surge Protection

Prevents power surges (e.g. from Defibrillation) to hit the amplifier

Prevents power from a faulty device to endanger the patient

Possible solutions:

• Clamping (here: Transistors start conducting at 0.2V to GND )

• Varistor: Massively increases resistance over clamping voltage

• Transient voltage suppression diode

• Thyristor: Can be created from a PNP- and NPN-Transistor

• … many more

https://commons.wikimedia.org/wiki/File:Typische_Varistorkennlinien.gifhttps://commons.wikimedia.org/wiki/File:Thyristor.svg

VaristorThyristor

10/30/2020

5


Galvanic insulation

No electric connection between two circuits

Possibilities:

• Induction: Two Transformers only connected by their magnetic field, most common, only AC

• Opto-isolator: LED + Photo-sensor, common for digital signals

• Relay: Switch a circuit by activating a magnet

https://commons.wikimedia.org/wiki/File:Trafo-innenleben.jpg https://commons.wikimedia.org/wiki/File:Optoisolator_topologies_both.svg

https://commons.wikimedia.org/wiki/File:Relay_principle_horizontal_new.gif


Instrumentation Amplifier

Consists of 3 Operational Amplifier

𝑉 = 𝑉 − 𝑉 · 𝐴 = 𝑉 − 𝑉 · 1 + ·

Advantages:

• Low noise

• Low drift

• High common-mode rejection

• High gain

https://commons.wikimedia.org/wiki/File:Op-Amp_Instrumentation_Amplifier.svg

10/30/2020

6


Noise Filter

EEG:

• Band Stop: 50Hz (common-mode interference, mains frequency)

ECG:


• High Pass: ~0.05Hz (Reduce breathing noise)

• Low Pass: ~150Hz (~45-60Hz for noisy signal, can only be used for rhythm analysis)

EMG:


• High Pass: ~20Hz

• Low Pass: ~400Hz


OpenEEG Filters

High-Pass: f = 0.16𝐻𝑧

Low-Pass: f = 59𝐻𝑧 (Besselworth = Combination of Bessel- & Butterworth-filter)

10/30/2020

7


Right leg driver

Optional, in addition to noise filter

Electricity from power socket has 50 or 60Hz

Electric devices or cables near patient create 50/60Hz interference

Use of RLD for active canceling of common-mode interference:

• Apply a known direct current to the body

• Better: current is dynamically adapted to common-mode signal

• Average leads, invert and halve, feed to body J.G. Webster, "Medical Instrumentation", 3rd ed, New York: John Wiley & Sons, 1998, ISBN 0-471-15368-0.


Analog-to-digital converter / AD Converter

Converts a current to a digital value

For Biopotentials needed:

• High resolution (e.g. 24bit)

• High Sampling rate

• At least double of the highest frequency signal

• ECG: 250Hz, EMG: 800Hz

Easy, cheap, solution: Sound card (16bit + 48kHz)

10/30/2020

8


Circuit to measure electric signals (recap)

1. Electrode

2. Surge protection / galvanic insulation

3. Instrumentation Amplifier

4. Noise Filtering

5. Analog-to-digital converter

Right leg driver to reduce noise

http://openeeg.sourceforge.net


ELECTROCARDIOGRAPHYECG

10/30/2020

9


ECG Short

Several (2/3/4/10) electrodes on torso & extremities

Measure electrical activity of the heart

Voltage: ~ 1-3mV

Time (Pulse ~80, shorter for faster pulse):

• p-Wave:

10/30/2020

10


ECG Leads

With 3 Electrodes (small ECG) 6 possible Leads:

• Defined by Einthoven:

• I: Right Arm (RA) → Left Arm (LA)

• II: RA → Left Leg (LL)

• III: LA → LL

• Defined by Golberger:

• aVF: RA+LA → LL

• aVL: RA+LL → LA

• aVR: LA+LL → RA

https://en.wikipedia.org/wiki/File:Limb_leads_of_EKG.png


ECG Leads

10 electrodes 12 leads:

• Einthoven + Goldberger leads +

• Wilson:

• Wilsons’s central terminal: LL+LA+RA

• V1: Wct → V1

• V2: Wct → V2

• …

Optional additional positions:

• V4R: Mirror position of V4

• V7-V8: Mirror positions of V4-V6 on the back

https://commons.wikimedia.org/wiki/File:De-Chest_leads_(CardioNetworks_ECGpedia).png

10/30/2020

11


ECG ARTIFACTS


Loose lead artifact

Artifacts in Lead I and II -> Electrode on right arm is loose

10/30/2020

12


Patient movement

Patient movement creates short bursts of high frequency noise

More often if electrodes are at end of extremities (Hand, Foot)

Tell Patient to stop moving


Baseline drift

Device has not found a stable mean voltage

Common after a new electrode is connected

10/30/2020

13


Shivering

Patient is cold and shivers, may look like atrial fibrillation


Pacemaker

Sharp spike (>100Hz) instead of p-Wave

©2015 Physio-Control, Inc., Redmond WA, USA. GDR 3306627_A

10/30/2020

14


Gastric/Neuro/… Stimulator

A lot of Spikes >40Hz

Some devices can be turned of with a magnet



Powerline

Constant 50 Hz noise

Use 50 Hz filter

Or move Patient/Cables away from Powerlines


10/30/2020

15


Reversed lead

ECG is reversed: Electrodes are switched

Correct placement

http://www.mauvila.com/ECG/ecg_artifact.htm


Chest compression

Artifacts created by chest compression during resuscitation (right side, Postshock)

Rhythm pre shock is a ventricular fibrillation

10/30/2020

16


ECG PATHOLOGIES


Some Sinus node Abnormalities

• Sinus Bradycardia, Sinus Tachycardia

• Too slow, too fast

• Premature atrial contraction

• There is another part of the atrium acting as a sinus node

• Two shapes for p-Waves and/or QRS

• Arrhythmic heart beat

• Multifocal atrial tachycardia

• Like PAC but many additional sinus nodes

https://commons.wikimedia.org/wiki/File:Multifocal_atrial_tachycardia_-_MAT.png

10/30/2020

17


Some Atrial Abnormalities

• Atrial Flutter

• Fast atrial activation, many p-Waves

• Only some are transmitted by the AV node

• Atrial Fibrillation

• chaotic, fast, activation of atria

• Irregular QRS complex

https://commons.wikimedia.org/wiki/File:Atrial_Flutter_Unlabeled.jpg

https://www.uni-heidelberg.de/presse/ruca/2010-2/3med.html


Some atrioventricular node abnormalities

• AV Node reentrant tachycardia

• Circular electric pathway at AV node

• AV node reactivates itself and atria

• AV Blocks:

• 1. Degree: longer delay

• 2. Degree Mobitz I: delay gets longer, until missing QRS

• 2. Degree Mobitz II: missing transmission

• 3. Degree: no relation between QRS and P-waves

https://commons.wikimedia.org/wiki/File:AV_nodal_reentrant_tachycardia.png

https://commons.wikimedia.org/wiki/File:Heart_block.png

10/30/2020

18


Some ventricular abnormalities

• Premature ventricular contraction “extra heart beat”

• Ventricular tachycardia

• Very fast but synchronized activation of ventricles

• Ventricular fibrillation

• Fast, chaotic activation of ventricles

• No heart beat -> No pulse

https://commons.wikimedia.org/wiki/File:PVC10.JPG

http://hqmeded-ecg.blogspot.com/2013/02/regular-wide-complex-tachycardia-what.htmlhttps://acls-algorithms.com/rhythms/ventricular-fibrillation/


THINGS ONE CAN DO WITH ECG

10/30/2020

19


Vectorcardiography

Display electric activity as position and direction

𝑋 = 0.172 𝑉1 + 0.074 𝑉2 − 0.122 𝑉3 − 0.231 𝑉4

−0.239 𝑉5 − 0.194 𝑉6 − 0.156 𝐼 + 0.010 𝐼𝐼

𝑌 = 0.057 𝑉1 − 0.019 𝑉2 − 0.106 𝑉3 − 0.022 𝑉4

+0.041 𝑉5 + 0.048 𝑉6 − 0.227 𝐼 + 0.887 𝐼𝐼

𝑍 = 0.229 𝑉1 + 0.310 𝑉2 + 0.246 𝑉3 + 0.063 𝑉4

−0.055 𝑉5 − 0.108 𝑉6 − 0.022 𝐼 − 0.102 𝐼𝐼

Yang, H., Bukkapatnam, S.T. & Komanduri, R. Spatiotemporal representation of cardiac vectorcardiogram (VCG) signals. BioMed Eng OnLine 11, 16 (2012). https://doi.org/10.1186/1475-925X-11-16


Automated rhythm analysis

Using deep learning/machine learning

Using statistics and decision trees:

• Frequency

• Rhythmic/Arhythmic

• Timing of PQ/QRS/ST ECG parts

Ribeiro, A.H., Ribeiro, M.H., Paixão, G.M.M. et al.Automatic diagnosis of the 12-lead ECG using a deep neural network. Nat Commun 11, 1760 (2020). https://doi.org/10.1038/s41467-020-15432-4

10/30/2020

20


How to compare Automated Rhythm analysis results

Ground truth defined by physicians

Compare algorithm results with statistics:

• Sensitivity (Correctly detected rhythm)

• Specificity (Detected absence of rhythm)

• True Positive & False Positive

• True Negative & False Negatives

• Positive predictive value PPV

• If algorithm detects a rhythm how likely is it that this rhythm is present

You want high Sensitivity/Specificity/PPVE. Manibardo et al., "ECG-based Random Forest Classifier for Cardiac Arrest Rhythms," 2019 41st Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), Berlin, Germany, 2019, pp. 1504-1508, doi: 10.1109/EMBC.2019.8857893.

Rhythm diagnosed by physician

Rhy

thm

dia

gno

sed

by a

lgo

rithm


Rhythm analysis using Deep Learning

Neural network to analyze ECG

Input is the data from 12 leads

• Doesn’t need to be this way. A neural network could also use the raw data from the electrodes

Outputs which rhythms the network has found

• Multiple diagnosis can be found

• Confidence for a diagnosis can also be returned

Zhu, H., Cheng, C., Yin, H., Li, X., Zuo, P., Ding, J., ... & Hu, S. (2020). Automatic multilabel electrocardiogram diagnosis of heart rhythm or conduction abnormalities with deep learning: a cohort study. The Lancet Digital Health.

10/30/2020

21


Automated rhythm analysis during CPR

Rhythm analysis during cardio pulmonary resuciation

Reduce no-flow time

Needs to remove Chest Compression artifacts

Fumagalli, F., Silver, A. E., Tan, Q., Zaidi, N., & Ristagno, G. (2018). Cardiac rhythm analysis during ongoing cardiopulmonary resuscitation using the Analysis During Compressions with Fast Reconfirmation technology. Heart rhythm, 15(2), 248-255.


Rhythm analysis using smartwatch / wearable

Acquire a single lead (usually I), also called iECG

• One arm wears the watch,

• the other touches an electrode on the watch

Can be used to detect arterial fibrillation

Also possible but not an ECG: Photoplethysmogram

• Heart rate and blood flow analysis

• Using LED + Photosensor

• Similar to Oximetry

Raja, J. M., Elsakr, C., Roman, S., Cave, B., Pour-Ghaz, I., Nanda, A., ... & Khouzam, R. N. (2019). Apple Watch, Wearables, and Heart Rhythm: where do we stand?. Annals of Translational Medicine, 7(17).

10/30/2020

22


ELECTROENCEPHALOGRAPHYEEG


EEG Short

Many (2/3/4/10) electrodes on the head

Measure electrical activity of brain parts

Voltage: ~ 5-100 μV (0.005-0.1 mV)

Signal separated by Frequency:

Delta Waves : 0.1-4 Hz

Theta Waves: 4-8 Hz

Alpha Waves: 8-13 Hz

Beta Waves: 13-30 Hz

Gamma Waves: >30 Hz

https://commons.wikimedia.org/wiki/File:EEG_cap.jpg

https://commons.wikimedia.org/wiki/File:ElectroEncephalogram.png

10/30/2020

23


EEG Waves / Frequencies

Average over all neurons in the area around the electrode

Signal is similar to 1/f noise (pink noise)

Division into frequency bands is purely historic

Physicians analyze by finding and interpret patterns

https://commons.wikimedia.org/wiki/File:Eeg_raw.svg


Delta Waves

Normal:

• Adults, front brain, during sleep (slow-wave sleep)

• Children, rear brain

Pathologic:

• Delta wave activity corresponds with sleep disorders

• Disruption of delta waves increases risk of diabetes Type II

• Schizophrenia reduces delta waves during sleep

• Parkinson’s show disrupted delta waves

https://commons.wikimedia.org/wiki/File:Eeg_delta.svg

10/30/2020

24


Theta Waves

Normal:

• In Cortical areas, during meditation, drowsiness, sleep

• Not present during deep sleep

• Spikes if a response/action is repressed

Regular Waves 4-10Hz in Hippocampus are also called (hippocampal) theta waves

• Waves are removed if medial septal area is damaged or inactivated

• Very regular

https://commons.wikimedia.org/wiki/File:Eeg_theta.svg


Alpha Waves

Normal:

• While awake Occipital lobe (rear brain, visual processing):

• Reduced with open eyes or during sleep

• During REM sleep, frontal-central

Pathological:

• Alpha wave intrusion during sleep, disrupts delta waves, sleep disorders

https://commons.wikimedia.org/wiki/File:Eeg_alpha.svg

10/30/2020

25


Beta Waves

Normal:

• During wake time

• Associated with thinking, concentrating

• At motor cortex: movement, higher if movement is suppressed

• Artificial stimulation slows movement

https://commons.wikimedia.org/wiki/File:Eeg_beta.svg


Gamma Waves

Normal:

• During wake time

High noise from (eye) muscles!

• Gamma waves correlate to eye movement (saccade)

Pathological:

• Depression or Bipolar disorder have altered gamma waves

• Schizophrenia reduces activity

https://commons.wikimedia.org/wiki/File:Eeg_gamma.svg

10/30/2020

26


Mu wave

Frequency: 7.5 – 12.5 Hz (overlaps with alpha waves)

Only long the motor cortex

Suppressed if mirror neurons are active

Interesting for Brain-computer interfaces:

• Mu waves can be changed by thinking about doing something

• Different locations correspond to different body parts

• Generally left brain -> right body, and the reverse


Sensorimotor rhythm waves (SMR Wave)

Frequency: 12.5 – 15.5 Hz (low beta waves)

Only along the motor cortex

Get stronger during motion, or imaging to do the motion

Might be used for Brain-Computer interfaces

• Useful for disabled people

10/30/2020

27


Brain-Computer Interface

Common Theme in Science Fiction:

• Matrix, Star Trek (Borg), Ghost in the Shell, Altered Carbon, RoboCop, Neuromancer, Pacific Rim, Chappie ...

• Control Computers like a third arm

In Science: still under development

Differentiation to Neuroprosthetics:

• Neuroprosthetics replace functions of the brain or other neural structures:

• Cochlear implants (Replaces inner ear), Retinal implants, Pacemaker

• Sometimes used interchangeably to Brain-Computer Interfaces


Invasive Brain-Computer Interface

Electrodes inside the brain

- Needs surgery

+ Better signal, Low noise

+ High spatial resolution

Examples:

• 1987 Human with acquired blindness sees light (shades of gray, very small field, low framerate)

• ~2002: Human with locked-in syndrome controls a mouse cursor

• 2002 Update of 1987 BCI, blind person could drive a car (slowly, on a parking lot)

• 2004 Dobelle, inventor of the BCI for blindness dies, the patients developed problems with their visions that nobody could fix. They are now blind again.

• 2005: Tetraplegic human controls a robot hand

10/30/2020

28


BrainGate

Invasive Brain-Computer Interface

Controls a robot arm

Tetraplegic person can grab a bottle and drink from it


Partially invasive Brain-Computer Interface

Inside the Skull, outside the brain

- Needs Surgery

+ Better Signal, low noise

+ High spatial resolution

+ Less risk of injury to the Brain

Examples:

• 2004: Teenager plays Space Invaders

10/30/2020

29


Non-invasive Brain-Computer Interface

Electrodes on the skin

- High noise, low signal strength

- More Artifacts from motion, moving electrodes, bad conductivity, hair

+ No surgery needed

+ Easy, fast installation, reusable, removable

Examples:

• 1977 Control of a mouse cursor

• 1988 Text input


Brain-Computer Interfaces

To Play at home:

• OpenEEG project (Seen on previous slides) http://openeeg.sourceforge.net/

• OpenBCI http://openbci.com/

• SmartphoneBCI https://icibici.github.io/smartphone-bci-hardware/

https://commons.wikimedia.org/wiki/File:UCM3.jpg

10/30/2020

30


Neurochips

Neurons grown on semiconductor (computer) chips

Depending on design:

• Possible to access each neuron individually, or

• High spatial resolution

Neurons are susceptible to bacteria, need sustenance

1997: First neurochip, 16 neurons.

2004: Aircraft simulation controlled by a neurochip with 25k neurons, grid of 60 electrodes.

https://thesis.library.caltech.edu/1399/2/thesis_je.pdf


ELECTRO*GRAPHYE*G

10/30/2020

31


Tongue

Electropalatography:

• Measure how the tongue is used for articulation

• Electrodes on

• Right: O = contact, . = no contact

• Articulation of the word “catkin” /kæt.kɪn/ (excerpt)

https://commons.wikimedia.org/wiki/File:Epg-frames.JPG

https://commons.wikimedia.org/wiki/File:Electropalate-vertical.jpg


Ear

Electrocochleography:

• Measure potentials generated in the inner ear (cochlea)

• Electrodes either invasive (needles), or on skin

• A short (100ms) broadband sound is played to the patient

• Record frequencies between 10Hz and 1.5kHz

10/30/2020

32


Eye

Electrooculography:

• Measure eye movement

• Pair of electrodes, above/below or left/right of eye

• Measured potentials changed depending on the eye position

Electronystagmography:

• Used to record nystagmus (involuntary eye movement)

• Electrodes are placed around the eye


Muscles

Electromyography

Electrodes on skin, or with needles inside the muscle

Normal:

• No electricity if muscle is not used

• More strength used -> more action potentials

• Fully Contracted muscle -> unordered appearance of potentials

• Muscle fatigue -> higher amplitudes, lower frequency, longer potential duration

10/30/2020

33


EMG nonmedical uses

Measure:

• Sense isometric muscle use (no visible movement)

• Muscles used for speech, but without creating actual sound

Use:

• Control devices with motionless gestures

• Speech control in noisy environments or for people with certain speech disorders

• Video Game Controller (Microsoft, Facebook)


Abdomen

Electrogastrogram or Electrogastroenterogram

Electrodes on the skin

Measure activity in waves per minute

Stomach: 0.03-0.07 Hz = 3 waves / min

Duodenum: 0.18-0.25 Hz = 12 waves / min

10/30/2020

34


Questions for this Lecture

Send me an email:

[email protected]

Phone:+49 (0) 621 383 4603

Or ask during the live lecture:

19.11.2020 13:00

bme2020 2 ecg - umm.uni-heidelberg.de&kulvwldq 7|qqhv , 6olgh , *doydqlf lqvxodwlrq 1r hohfwulf...

Documents