Basic Biosignal Manipulation Full

BASIC BIOSIGNAL MANIPULATION
John Fredy Ochoa Gómez

Bioengineering Department
University of Antioquia
Biomedical Signal Processing 2
CONTENT
1. Introduction
2. File formats
3. Montage
4. Signal manipulation
5. Basic statistics
BASIC EEG ANALYSIS

Introduction
• In this part of the course we need to use the data commonly
acquired by the biomedical instrumentation in a format
commonly use in a data science environment using libraries
commonly used in the industry
Acquisition Manipulation Analysis
PYTHON DATA
SCIENCE
FRAMEWORK
NUMPY
PANDAS
MATPLOTLIB
CONTENT
1. Introduction
2. File formats
3. Montage
5. Basic statistics
BASIC EEG ANALYSIS

File formats
• Different to medical imaging, there are so many formats for
Biosignals.
• The initiative to establish the EDF or Biosig as standard file

format doesn’t has the force that has the DICOM format
• All the Biosignals files format are structured to have

information about the number of channels and the sampling
rate
• The following slides will show different formats and examples,

existing different possibilities, to load in Python
BASIC EEG ANALYSIS

File formats
• Matlab format: the scipy library is used to load the information
in the .mat file. The information loaded is assigned to a
dictionary variable where is necessary to extract the field with
the signal and related information
BASIC EEG ANALYSIS

File formats
• Otherwise, for formats as OpenBCI Raw we have plain text
Information
•
The file is commonly know as a comma separated values
(csv)
BASIC EEG ANALYSIS

File formats
• The following code is a propose of algorithm to read OpenBCI
Raw format
• First, some variables

BASIC EEG ANALYSIS

File formats
• The following code is a propose of algorithm to read OpenBCI
Raw format
• Now, the algorithm to load the data
BASIC EEG ANALYSIS

To solve
• Write the code in the preceding slides
• For the eight EEG channels loaded, obtain a plot similar, not
equal because the picture is from the web, to:
• To obtain the space between points remember that each

second 250 samples are acquired
BASIC EEG ANALYSIS

To solve
• Evaluate a strategy to load the data using the PANDAS
routines (see also: https://www.shanelynn.ie/python-pandas-
read_csv-load-data-from-csv-files/ )
CONTENT
1. Introduction
2. File formats
3. Montage
5. Basic statistics
BASIC EEG ANALYSIS

Montage / Visualization
• Given that is impossible to have many sensor a common
practice is to locate some electrodes using position that can
be compared between subjects or center. The location, of the
active and the reference electrode, is known as montage
• In the picture the common location of EEG electrodes
BASIC EEG ANALYSIS

• For the OpenBCI we have used the following electrodes
BASIC EEG ANALYSIS

• OpenBCI
BASIC EEG ANALYSIS

• The signal used is the difference between active and passive or
reference electrodes. So, for the loaded signal we have:
BASIC EEG ANALYSIS

To solve
• Plot the six biosignal obtained in the previous slide using a
subplot of 3x2 and assign the corresponding title, and the
titles of the axis, to each plot
• To obtain the space between points remember that each second 250
samples are acquired
CONTENT
1. Introduction
2. File formats
3. Montage
5. Basic statistics
BASIC EEG ANALYSIS

Signal manipulation
• For the following examples we will load a pre-processed
signal
BASIC EEG ANALYSIS

Signal manipulation
• This signal is divided in trials, also named epochs or
segments, where each trial is composed of points
• So, we have: 8 channels, 2000 points (for trial), and 180 trials
BASIC EEG ANALYSIS

To solve
• For the signal, plot the trial 120, channel 2

• Plot all the channels, separated, trial 100
• Plot the first channel, first epoch
• To obtain the space between points remember that each second 1000
samples are acquired
BASIC EEG ANALYSIS

Signal manipulation
• This signal is divided in trials, also named epochs or
segments, where each trial is composed of points
• So, we have: 8 channels, 2000 points (for trial), and 180 trials
To solve
• How we can change the shape of the signal to obtain a new

signal with dimensions: 8 channels, 2000 x 180 points?
• Use the numpy’s method reshape and evaluate the different options, A,
B, C, F, comparing the signal obtained (compare with the first channel
first epoch of the segmented signal)
BASIC EEG ANALYSIS

To solve
• For a continuous signal develop an algorithm in Python to test
if the signal is random using the following theory
• Test for randomness: Random signals are generally expected to display more
excursions about a certain reference level within a specified interval than signals that are
predictable.
• A peak or a trough is defined by a set of three consecutive samples of the signal, with the
central sample being either the maximum or minimum, respectively.
• As the direction of excursion of the signal changes at peaks and troughs, such points are
collectively known as turning points.
• A simple test for a turning point is that the sign of the first-order difference (derivative) at the current
sample of the signal be not equal to that at the preceding sample.
• Given a signal of N samples, the signal may be labeled as being random if the number of
turning points is greater than the threshold ( − 2)
BASIC EEG ANALYSIS

To solve
• For a continuous signal develop an algorithm in Python to test
if the signal is random using the following theory
BASIC EEG ANALYSIS

Signal manipulation
• The division of the signal in epochs is a common approach in
exams where a stimuli or condition is frequently repeated
BASIC EEG ANALYSIS

To solve
• Average all the epochs for each signal to obtain two new
signals with dimensions 58x1150
• Construct a subplot of 5x4, in each entry plot one of the last

20 channels for the average of epochs for
condicion_frecuente and condicion_infrecuente
• To obtain the space between points remember that each second 1000 samples
are acquired
CONTENT
1. Introduction
2. File formats
3. Montage
5. Basic statistics
28
BASIC EEG ANALYSIS
• A deterministic signal is one whose value at a given instant of

time may be computed using a closed-form mathematical
function of time, or predicted from a knowledge of a few past
values of the signal.
• A signal that does not meet these condition may be labeled

as a nondeterministic signal or a random signal.
• Most of the physiological signals (EEG, ECG, blood flow,

human gait, etc.) are characterize by complex dynamics
including both non-stationarities and non-linearities.
29
BASIC EEG ANALYSIS
• Although not necessary, lets introduce some DataFrame

manipulation
30
BASIC EEG ANALYSIS
To solve
Use the method pandas.DataFrame.describe and discuss the

obtained values
https://pandas.pydata.org/pandas-docs/stable/reference/api/pandas.DataFrame.describe.html
31
BASIC EEG ANALYSIS
Types of Data
There are two main types of data: qualitative or categorical

data and quantitative or numerical data.
• Qualitative: non-numerical information such as race, gender,

marital status, etc.
• Quantitative: measurement or count such as age, salary,

height, weight, etc.
32
BASIC EEG ANALYSIS
Basic DataFrame description

All the data are numeric
33
BASIC EEG ANALYSIS
Centrality measures
One common way to summarize our numerical data is to find

out the central tendency of our data.
“What is the most typical value of the voltage in our dataset?”.

To address this question, we can use common measures of
center: mean and median.
To solve
Which is the difference between mean and median?

34
BASIC EEG ANALYSIS
Centrality measures
35
BASIC EEG ANALYSIS
Variability measures
Variation is always observed in a dataset.
The Standard Deviation (SD) is a measurement of how a set of

values spread out from their mean.
• SD is a positive number
• SD is affected by outliers
• The smallest possible value of SD is zero.
• If SD is zero, all the numbers in a dataset have the same value.
36
BASIC EEG ANALYSIS
Variability measures
37
BASIC EEG ANALYSIS
Graphical inspection
Boxplot
One approach to reveal the data distribution is to find a Five-
Number Summary from our dataset. The Five-Number
Summary includes:
• The minimum
• The 25th percentile or the first quartile (Q1)
• The median
• The 75th percentile or the third quartile (Q3)
• The maximum
To solve: What is the definition of percentile?

38
BASIC EEG ANALYSIS
Boxplot
39
BASIC EEG ANALYSIS
Boxplot
40
BASIC EEG ANALYSIS
Histogram
Is a graphical display that uses rectangular bars to show the
frequency distribution of a set of numerical data.
41
BASIC EEG ANALYSIS
Histogram
42
BASIC EEG ANALYSIS
To solve
Explore, and show examples, of others alternatives to plot and

describe the data in the DataFrame
43
BASIC EEG ANALYSIS

TEST FOR STATIONARITY
- Look at plots: scatter, histogram
- Summary statistics: mean, variance, etc
- Statistical tests: Augmented Dickey-Fuller Test

- H0: The time-series is non-stationary
- H1: The time-series is stationary
44
BASIC EEG ANALYSIS Testing procedures for non-stationarity and non-linearity in physiological signals

• The first simple test may be to observe the amplitude distribution.
• In most of the methods for data processing based on stochastic models,

gaussian distribution is assumed.
45
BASIC EEG ANALYSIS

Testing procedures for non-stationarity and non-linearity in physiological signals
46
BASIC EEG ANALYSISTesting procedures for non-stationarity and non-linearity in physiological signals
• The histograms are constructed for the first and second half of data
separately and obvious non-gaussianity could be seen during the task
performance.
• Then, we have a non-stationary signal

47
BASIC EEG ANALYSIS Testing procedures for non-stationarity and non-linearity in physiological signals
• The second test is based on inspection of the shape of the autocorrelation

function (ACF).
• The slow decrease of ACF for large lags indicates long-range correlation,
which may be due to non-stationarity and/or dynamic non-linearity.
• In case of non-stationarity due to additive (e.g. trend-like) components,

they could be removed in order to test the ACF on the residuals.
• If the shape of ACF remains the same, this may be a typical case of non-
stationarity.
48
BASIC EEG ANALYSIS


49
BASIC EEG ANALYSIS

• The possible non-linearity may be tested by the comparison of the shape

of square of ACF with that of ACF of squares of samples.
• If ( )≠( ( )) , non-linear process is suggested.
• Fig. 2 shows the plots of ACF computed for the part of the same EEG.
• The long-range correlation and different shapes of ( ) and

( ( )) suggest both non-stationarity and non-linear dynamics.
50
BASIC EEG ANALYSIS

51
BASIC EEG ANALYSIS

To solve
• Discuss, using all the posible graphs and data previously discussed, the
stationarity and nonlinearity in the signals in the DataFrame

Basic Biosignal Manipulation Full

Uploaded by

Copyright:

Available Formats

Basic Biosignal Manipulation Full

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Basic Biosignal Manipulation Full

Uploaded by

Copyright:

Available Formats

BASIC BIOSIGNAL MANIPULATION

John Fredy Ochoa Gómez

BASIC EEG ANALYSIS

BASIC EEG ANALYSIS

• The initiative to establish the EDF or Biosig as standard file

• All the Biosignals files format are structured to have

• The following slides will show different formats and examples,

BASIC EEG ANALYSIS

BASIC EEG ANALYSIS

BASIC EEG ANALYSIS

• First, some variables

BASIC EEG ANALYSIS

BASIC EEG ANALYSIS

• To obtain the space between points remember that each

BASIC EEG ANALYSIS

BASIC EEG ANALYSIS

BASIC EEG ANALYSIS

BASIC EEG ANALYSIS

BASIC EEG ANALYSIS

BASIC EEG ANALYSIS

BASIC EEG ANALYSIS

BASIC EEG ANALYSIS

BASIC EEG ANALYSIS

• For the signal, plot the trial 120, channel 2

BASIC EEG ANALYSIS

• How we can change the shape of the signal to obtain a new

BASIC EEG ANALYSIS

BASIC EEG ANALYSIS

BASIC EEG ANALYSIS

BASIC EEG ANALYSIS

• Construct a subplot of 5x4, in each entry plot one of the last

BASIC EEG ANALYSIS

• A deterministic signal is one whose value at a given instant of

• A signal that does not meet these condition may be labeled

• Most of the physiological signals (EEG, ECG, blood flow,

BASIC EEG ANALYSIS

• Although not necessary, lets introduce some DataFrame

BASIC EEG ANALYSIS

Use the method pandas.DataFrame.describe and discuss the

BASIC EEG ANALYSIS

There are two main types of data: qualitative or categorical

• Qualitative: non-numerical information such as race, gender,

• Quantitative: measurement or count such as age, salary,

BASIC EEG ANALYSIS

Basic DataFrame description

BASIC EEG ANALYSIS

One common way to summarize our numerical data is to find

“What is the most typical value of the voltage in our dataset?”.

Which is the difference between mean and median?

BASIC EEG ANALYSIS

BASIC EEG ANALYSIS

Variation is always observed in a dataset.

The Standard Deviation (SD) is a measurement of how a set of

BASIC EEG ANALYSIS

BASIC EEG ANALYSIS

To solve: What is the definition of percentile?

BASIC EEG ANALYSIS

BASIC EEG ANALYSIS