Biosignal processing:

An introduction

Dr Mohamed Seghier
Goal
➔ to develop specific skills related to the acquisition and
processing of biosignals to extract a priori desired
information and interpreting the nature of a
physiological process based on the observations of how
the process alters the characteristics of a biosignal.
Course Learning Outcomes

1. Enhance skills for quantifying and (if necessary) compensating for

the effects of measuring devices and noise on a signal being measured.
2. Develop skills for identifying and separating desired and
unwanted components of a bio signal.
3. Enhance skills aimed at uncovering the nature of phenomena
responsible for generating the biosignal based on the identification and
interpretation of an appropriate model for the signal.
4. Understand how to relate the properties of a physical system to
the differences in characteristics between a signal used to excite (or
probe) the system and the measured response of the system.
Introduction
The living organism made up of many component systems and each system is made up
of several subsystems that carry on many physiological processes.

➔ Most physiological processes are accompanied by or manifest themselves as signals

that reflect their nature and activities.

Diseases or defects in a biological system cause alteration its

normal physiological processes, leading to pathological processes.
A pathological process is typically associated with signals that
are different in some respects from the corresponding normal
signals.
When something in the brain goes wrong
→ hundreds of neurological diseases and disorders.

Cost per year: World: $2.5 trillion; Europe: €798 billion; U.S.: $760 billion. [WHO, 2010]
Signals:
A signal is a function that conveys information.
Signals can be biochemical, electrical, or/and physical in nature

Biomedical signals are signatures of physiological activities that are expressed at

different spatial and temporal dimensions (e.g. gene and protein sequences,
neuronal activations and cardiac rhythms, images of tissue and organs…etc).

The signals can be sensed by qualitative or quantitative manner:

• Measurement • Scalar • Function of time • discrete x[n] • continuous x(f)
• digital • Multivariant vector
Example: Measurements of
the blood pressure of a
patient
Example: An electrocardiogram (ECG or EKG) records the electrical signal
from your heart to check for different heart conditions.
Example: ECG in a patient with an anterolateral myocardial infarction (MI)

https://www.emra.org/emresident/article/acute-mi-case-report/
Example: a normal Electroencephalography (EEG) signal in a 16-month-old infant.
Example: EEG in a 32-year-old man with right
mesial temporal lobe epilepsy
https://www.ncbi.nlm.nih.gov/books/NBK390347/#pt7.s1
Example: Schizophrenia patients have abnormal neural
responses to salient, infrequent events (ERP signals)

https://www.sciencedirect.com/science/article/pii/S2213158215001230
Example: Speech signal of the word “safety” uttered by a male speaker
Segments of the signal on an expanded scale to illustrate the quasi-periodic nature of the voiced sound /E/ in the upper
trace, and the almost-random nature of the fricative /S/ in the lower trace
Example: diagnosis of Parkinson’s Disease (PD)

Age-standardised prevalence of Parkinson's disease per 100 000 population by location for both sexes, 2016
https://www.thelancet.com/journals/laneur/article/PIIS1474-4422(18)30295-3/fulltext
https://www.sciencedirect.com/science/article/pii/S0896627303005683
https://www.aans.org/en/Patients/Neurosurgical-Conditions-and-Treatments/Parkinsons-Disease
How to diagnose disorders over the phone
• Subjects speak over the phone.
• A system will analysis the features of their speech.
• Make a diagnosis about potential conditions/disorders

https://www.ox.ac.uk/news/2015-06-18-how-
diagnose-disorders-over-phone
Challenge:
Diagnosis of COVID-19 over the phone? Possible?
➔ Real-world behavior results from interactions among sources of energy and
modifiers of energy.

Signal processing addresses:

1- the inferences that once can draw from a signal about the property of
the source
2- the relationships among simultaneously observed signals.

As engineers, you can:

• Measure signals
• Manipulate signals
• Describe qualities of the source
• Probe the source
Objectives of Biomedical Signal Analysis:
• Information gathering: measurement of phenomena to interpret a system.
• Diagnosis: detection of malfunction, pathology, or abnormality
• Monitoring: obtaining continuous or periodic information about a system
• Therapy and control: Modification of the behavior of system based upon
the outcome of the activities listed above to ensure a specific result
• Evaluation: Objective analysis to determinate the ability to meet functional
requirements, obtain a proof of performance, perform quality control, or
qualify the effect of treatment
• Prognosis: ability to predict outcome after disease or treatment
Acquisition procedures:
Invasive: placement of transducers or other devices inside the body

Noninvasive: acquisition with no or minimal risk (e.g. surface electrodes)

Active: require external stimuli

Passive: does not require external stimuli

Invasive Noninvasive
Electroencephalogram (EEG)
Intraoperative electrocortical stimulation
Active Passive

An ECG during a physical exercise An electrocardiogram (ECG or EKG)

Properties of Biomedical Instruments
• Isolation of the subject or patient, or interaction with tools and environment.
• Range of operation: The minimum to maximum values of the signal being measured.
• Sensitivity: the smallest signal variation measurable.
• Specificity: the extent to which a measurement is specific for a particular condition, trait or
parameter.
• Resolution (temporal or/and spatial): a measure of the amount of detail in a signal.
• Linearity: consistency of measurements over the entire range of measurements.
• Hysteresis: a lag in measurement due to the direction of variation of the entity being
measured.
• Frequency response: represents of the variation of the sensitivity with frequency.
• Stability: an unstable system could preclude repeatability and consistency of
measurements.
• Signal to noise Ratio: noises could compromise the quality of the signal being acquired.
• Accuracy: the closeness of the measured value to a standard or true value.
• Precision: the closeness of two or more measurements to each other (repeatability and
reproducibility).
Difficulties of Biomedical Signal Processing
• Accessibility of the variables to measurement
• Variability of the signal source (e.g. within- or between-subject variability)
• Inter-relationship and interactions among physiological systems
• Effect of the instrumentation or procedure on the system
• Physiological artifacts and interference
• Poor signal to noise ratio
• Multimodality (e.g. differences between acquisition modalities)
• Energy limitation
• Patient safety
Can measurement conditions impact upon collected data?

Is it possible to design an experiment to collect data with no noise?

Accuracy of measurement in the light of inter-subject variability in

physiological signals?
Systemic errors
A signal is a single-valued representation of information as a function of an
independent variable (e.g. time).
→ information: some form of energy produced by the system.
→ Information: may have real or complex values.

Signal can be a continuous-time x(t) or a discrete-time signal x[t].

A signal can be:

Deterministic: a signal whose future values can be exactly predicted if past information is
available (e.g. a sine wave or a signal expressed exactly in a mathematical form);

Stochastic: a signal for which it is impossible to predict future values even if one knows its
entire past history (e.g. random fluctuations and noise).

Fractal: a scale-invariant signal that looks similar at all levels of magnification (e.g part of
the beat-to-beat heart rate, current through a single ion-channel of a cell membrane).

Chaotic: a deterministic signal that cannot be predicted exactly in the future (predictable
for a short time but prediction errors becomes large in the future; e.g. some EEG have
some chaotic characteristics).
Motivation of signal processing:
• Remove unwanted signal components that are corrupting the signal of
interest
→example: filtering and removing noise (what is signal/noise?)

• Extract useful information by rendering it in a more obvious form

→example: discriminate between normal and abnormal signals for diagnostic
purpose.

• Predict future values to anticipate the behavior of its source

→ example: predict the onset of a disease, control the behavior of an
intervention
Example: Chest EMG during two breaths contaminated by the
ECG signal (top: original signal, bottom: detect and remove the
ECG signal
Example: Normal vs. abnormal signal

MR-spectroscopy water-suppressed proton

spectra of an 8-mL voxel located in the
parietal region including predominantly
normal-appearing white matter in a patient
with cirrhosis before (left) and after (right)
liver transplantation. The main resonances
correspond to N-acetylaspartate (NAA, 2.0
ppm), glutamine/glutamate (Glx, 2.1–2.5
ppm), creatine/phosphocreatine (Cr, 3.02
ppm), choline-containing compounds (Cho,
3.2 ppm), and myo-inositol (Ins, 3.55 ppm).
The initial spectrum shows an increase in
the glutamate/glutamine region and a
decrease in the myo-inositol and choline
resonances. These abnormalities
normalized after liver transplantation.

http://www.ajnr.org/content/29/9/1612
Example: Alzheimer-signature MRI biomarker predicts AD dementia in
cognitively normal adults

https://n.neurology.org/content/76/16/1395
Linguistic markers predict onset of Alzheimer's disease

https://www.thelancet.com/journals/eclinm/article/PIIS2589-
5370(20)30327-8/fulltext

➔ to use classification methods to predict future onset of

Alzheimer's disease in cognitively normal subjects through
automated linguistic analysis.

MRI accurately measures cortical thickness