Difference between revisions of "Paradigm File Format in BESA"

Revision as of 15:35, 8 April 2016

General remarks

A paradigm description file (“PDG file”) contains the information which is relevant for describing an experimental setup in terms of the stimulation and response events that occurred. To describe the experimental paradigm, three terms are introduced:

1. An attribute is used to group trigger events into a certain class. For example, in an auditory experiment, triggers could be grouped according to an attribute “modality” to distinguish stimulation and response, and another attribute “side” to distinguish left and right.
2. An attribute value defines how a trigger event is classified in the class defined by an attribute. For example, the trigger with the code 1 could be a tone stimulus, and the trigger with the code 2 could be the subject’s response. That means that for the “modality” attribute, trigger 1 would receive an attribute value “tone”, whereas trigger 2 would receive an attribute value “response”.
3. A condition defines which trigger events form the set of events that should be averaged. For example, this could simply be all trigger events with the modality “tone”, or all trigger events which have the modality “response”, and follow a trigger event with the modality “tone”.

The PDG file is written in ASCII format and can thus be viewed and edited in any text editor. It is subdivided into a maximum of 9 sections:

1. [Attributes] Attributes which are used to group the trigger events. This section is mandatory.
2. [Values] A table of attribute values which are defined for the triggers used in the experiment.
3. [Names] Names of conditions which are defined in detail in section 8
4. [Epochs] For each condition, averaging epochs, baseline epochs, and some other epochs are defined
5. [Thresholds] Threshold settings used for artifact rejection
6. [Averaging] Defines which conditions are selected for averaging
7. [Filter] Filter settings for averaging
8. [TimeFrequency] Settings for time-frequency analysis
9. [Selections] Each condition is written here as a statement using Boolean logic
10. [ArtifactScan] The artifact scan results are written here

Most sections can be omitted; their entries are then filled with default values. The default settings are given in the detailed description (see below). The most relevant sections for describing an experiment are Attributes, Values, and Selections.

Connection to BESA paradigm editing tool: Sections 1 and 2 define entries in the “Trigger” tab. Sections 3 and 9 define entries in the “Condition” tab. Section 4 defines entries in the “Epoch” tab. Sections 5 and 10 define entries in the “Artifact” tab. Section 6 defines entries in the “Average” tab. Section 7 defines entries in the “Filter” tab. Section 8 defines entries in the “Coherence” tab (if available).

All values can be edited in the respective tabs of the paradigm editing tool.

Detailed description of the sections

Attributes

This section is mandatory. It holds the attributes which are used to group the trigger events. The first attribute is always the trigger code (“code”). Attributes are separated by a tabulator. By default, the ERP module defines a second attribute “name” which can take trigger names as attribute values.

Example:

[Attributes]
code	name	modality

Values

This section holds a table of attribute values which are defined for the triggers used in the experiment. Each row defines one trigger. The first column holds the trigger codes, since the trigger code is the only mandatory attribute. The second column holds the values for the second attribute (default in the ERP module: name), and so on. If no value is defined for an attribute, “NULL” is entered. For trigger codes which are not listed, no values are defined.

Example:

[Values]
0	NULL	NULL
1	tone	auditory
2	rare	auditory
3	frequent	auditory
128	response	motor

The trigger with code 0 has no values defined, the trigger with code 1 has the name “tone” and the modality “auditory”, and so on.

Names

This section holds the names of conditions which are defined in detail in the section “Selections”. It is not necessary to give names to conditions. By default, conditions are simply named “Condition 1” through “Condition 32”. If a name is given, it is followed by the zero-based index of the condition it refers to. This enables mixing of conditions with user-given names and conditions with default names.

Example:

[Names]
target	0
hit	1

The first condition obtains the name “target”, the second one obtains the name “hit”.

Epochs

This section holds the averaging epochs for each condition. Furthermore, additional epochs can be specified. Each epoch is defined by a floating point value which gives the pre-stimulus interval in milliseconds (ms), and a second floating point value which gives the post-stimulus interval in ms. Epochs are given in the following order:

Averaging epoch: The epoch used for averaging

Baseline epoch: The epoch over which the baseline is calculated

Epoch used for artifact rejection: The epoch in which the data are checked for artifacts

Stimulus artifact epoch: An epoch where a stimulus artifact occurred. This epoch is interpolated over before artifact rejection takes place

Stimulus delay: Only one value. This value gives the delay of the stimulus relative to the trigger event (e.g. when tactile stimulation is performed, where air pressure builds up after the trigger was set).

Not all epochs have to be provided. If the epochs list for a condition is incomplete, default values are used.

By default, the following values are used:

Example:

[Epochs]
-500.0	1000.0	-100.0	0.0	-500.0	1000.0	0.0	0.0	0.0
-800.0	500.0	-800.0	-700.0	-800.0	500.0	0.0	0.0	0.0

The first condition is averaged from 500ms before the trigger event to 1000ms after the trigger event. Baseline is calculated in the pre-stimulus interval between 100ms before the trigger event and the trigger event. Artifacts are rejected between –500ms and +1000ms, i.e. over the entire averaging epoch. No stimulus artifact interval and no stimulus delay is given.

For the second condition, averaging and artifact rejection are set to –800...500 ms, and baseline calculation is performed in the interval from –800ms to –700ms. Again, no stimulus artifact interval, and no stimulus delay is given.

Thresholds

This section holds values for artifact rejection. It can hold two sets of values. The first set contains settings for the color normalization of the BESA artifact scan tool. These values are not directly used for artifact rejection, but merely for visual control if the BESA artifact scan tool is used for artifact rejection.

The line following the first set of values contains either the string “AUTO_REJECT”, or “MANUAL_REJECT”. The second set contains settings for automatic rejection of artifacts. It is used if the preceding line contains the string “AUTO_REJECT”. Otherwise, only artifacts which were manually marked in the data are excluded from the averaging process.

The following table gives the order of the entries in the thresholds section:

If no values are given , the following values are used by default:

The unit ∂T stands for the sampling interval.

Example:

[Thresholds]
100.0	1000.0
0.001	64.000
75.0	800.0
AUTO_REJECT
100.0	1000.0
0.001	64.000
75.0	800.0

Automatic artifact rejection is active. The rejection thresholds are:

• 100 µV for EEG amplitude,
• 1000 fT for MEG amplitude,
• µV/∂T for the square root of the variance of the gradient of the EEG signal,
• 64 fT/∂T for the square root of the variance of the gradient of the MEG signal,
• 75 µV/∂T for the gradient of the EEG amplitude, and
• 800 fT/∂T for the gradient of the MEG amplitude.

The display thresholds for the BESA artifact scan tool are set to the same values.

Rejection behavior:

A sweep is rejected from averaging if:

• The signal exceeds the max. amplitude threshold at any sampling point in any of the “good” channels.
• The variance of the gradient of the signal, taken over the whole sweep, does not exceed the variance threshold in any of the “good” channels. This criteria is especially useful for MEG channels, where channels may show a flat signal temporarily during the measurement.
• The gradient of the signal exceeds the max. gradient threshold at any sampling point in any of the “good” channels.

Averaging

This section holds the conditions which are selected for averaging, and further information about whether all sweeps or a sub-set will be averaged.

The first line holds the number of conditions currently selected (N). If N=0 or if the section is not present, all conditions which have at least 1 matching trigger event are automatically selected for averaging.

The next N lines are organized as follows:

Column 1: 1 if the condition is selected and activated, 0 if it is selected, but not activated.

Column 2: Zero-based index of the condition.

Column 3: Information about whether all sweeps or a sub-set will be averaged, encoded bit-wise:

If more than one bit is set, a different averaging buffer will be created for each sub-set.

Example:

[Averaging]
1
1	1	9

One condition was selected. The second condition (condition index 1) is selected and activated, and two buffers will be averaged: the first buffer for all matching sweeps (bit 0 set), the second buffer for the first half of matching sweeps (bit 3 set).

Filter

This section holds filter settings which were chosen for averaging. The following table gives the possible settings.

The parameters Slope and Type are set to one of the following values:

If any of the required settings are missing, the filters that are currently set in the data file are used for averaging.

Example:

[Filter]
0.200000	0	1	1
75.000000	1	0	0
0.000000	1.000000	FALSE
75.000000	7.500000	FALSE
FALSE	FALSE
FALSE	FALSE	FALSE
TRUE		FALSE

The filters are set as follows:

High pass at 0.2 Hz, 6 db/Oct, forward filtering, activated.

Low pass at 75.0 Hz, 12 db/Oct, zero phase filtering, not activated

Notch filter at 0 Hz, width 1 Hz, not activated

Band pass filter at 75 Hz, width 7.5 Hz, not activated

Polygraphic channels are filtered like the EEG and MEG channels. They are not rectified.

Selected channels are not differentiated, not rectified, and not smoothed.

High pass filter is also activated for the artifact scan, and low pass filter is not activated for the artifact scan.

So, the only filter that will be applied to the data for both averaging and artifact scan is the high pass filter at 0.2 Hz.

The last line holds activation settings for the artifact scan. Thus, it is possible to activate or de-activate filters solely for the artifact scan.

TimeFrequency

This section holds the settings of the Time-Frequency dialog. It is only read if the time-frequency module is loaded. The following values are stored:

Example:

[TimeFrequency]
2	6	0
25000.000000
2	FALSE	-1
0	TRUE
4	4
3	4.000000	6
1	0

In this example, the values have the following meaning:

1st line: Coherence analysis with any reference channel possible.

2nd line: The reduced sampling interval after Time-frequency transformation is 25 ms, i.e. 25000 µs.

3rd line: Target condition has condition index 2, no control condition was specified.

4th line: For regional sources, the radial orientation is taken. The Gaussian FIR filter is used.

5th line: Array entry with index 4 was chosen for the frequency spacing. The demodulation filter is calculated for a bandwidth of 4 frequency spacings.

6th line: Array entry with index 3 was chosen for the lower frequency cutoff. The lower cutoff frequency is 4 Hz. Array entry with index 6 was chosen for the higher frequency cutoff.

7th line: (currently not used) For mean coherence analysis, the whole data set is used. The epoch length of each epoch is given by array index 0.

Selections

This section holds the definition of conditions in boolean logic. Each line corresponds to either a logic statement, a bracket, or a boolean operator. Conditions are separated by blank lines. There are no default settings for this section.

Tokens are written in uppercase.

A statement consists of

• a qualifier that can be either CURRENT, PREVIOUS, or NEXT depending on which trigger event is tested
• an attribute (e.g. code, name, interval, ...)
• a comparison operator (IS, IS NOT, IS MORE THAN, IS LESS THAN)
• a value (e.g. “tone”, 300ms, ...)

Operators can be AND, OR. Bracketing is also possible.

Example:

[Selections]
CURRENT.name IS "tone"
 
CURRENT.name IS "response"
AND
PREVIOUS.name IS "tone"
AND
PREVIOUS.interval IS LESS THAN 300ms
 
CURRENT.name IS "response"
AND
(
PREVIOUS.name IS "rare"
OR
PREVIOUS.name IS "frequent"
)

This example contains 3 conditions.

The first one selects all trigger events with the value “tone” for the attribute “name”.

The second one selects all trigger events with the value “response” for the attribute “name” which are preceded by trigger events with the name “tone” with an interval between the two of less than 300ms.

The third one selects all trigger events with the name “response” which are preceded by either a trigger event with the name “rare” or a trigger event with the name “frequent”. The bracket is important; without the bracket, the order of testing is changed, with a possibly different result.

Artifact Scan

This section holds the results of an artifact scan. They are plotted in the artifact tab central window. For each epoch which was scanned, all values are written. The format contains 4 header lines, followed by 4 lines for each epoch:

Line 1: <Number of EEG channels> <Number of magnetometer or axial gradiometer channels> <Number of planar gradiometer channels>

Line 2: <Channel badness flag for channel 1> <Channel badness flag for channel 2> … <Channel badness flag for channel N>

The channel badness flag is set as follows:

Line 3: <Number of conditions><index of first condition in list><index of 2nd condition in list>…<index of last condition in list>

The list is the artifact scan list of conditions.

Line 4: <Start of scanned range (position in microseconds)><end of scanned range (position in microseconds>

The remaining lines come in groups of 4. Each group contains the settings for one epoch. For each epoch:

Line 1: <artifact scan list index><position in microseconds>

The artifact scan list index is not necessarily the same as the condition index. It is possible that no artifact scan list index exists for a condition index, e.g. if no matches were found for a condition. The artifact scan list index runs from 0 to the number of conditions for which a scan was performed. The relation to the real condition index is given by the header line 3.

Line 2: <max_amp_in_channel1>< max_amp_in_channel2>…< max_amp_in_channelN>

Line 3: <variance_in_channel1><variance_in_channel2>…<variance_in_channelN>

Line 4: <max_gradient_in_channel1>< max_gradient_in_channel2>…< max_gradient_in_channelN>