fMRI Fear Conditioning Paradigm - PennBBL/conte GitHub Wiki
Motivation:
To detail the fMRI Fear Conditioning paradigm (version3/design3 unless otherwise specified) that was created and run for CONTE, both run1 and run2 (reversal). Behavioral Ratings are detailed in the wiki "fMRI Fear Conditioning Ratings".
Task Design Overview:
During the run1 conditioning run, participants were conditioned using the paradigm described below. Fear conditioning was achieved in CONTE using a novel social aversion fMRI task where neutral male faces were paired with either no tone, creating a neutrally-conditioned stimulus (CS-), or with an aversive scream with a 50% contingency creating an aversively-conditioned stimulus (CS+) with both trials where the scream was present (CS+paired; CS+p) or not present (CS+unpaired; CS+u). The face stimulus is present on the screen for 6 total seconds. The scream that occurs during the CS+p trials is present for the last second of the trial so that the face is on for 5 seconds alone, then for 1 second concurrently with the scream. There were 24 CS- trials, 12 CS+p trials, and 12 CS+u trials in each run. Each run also includes "catch trials" or trials where the cross-hair will turn green on the screen and the participant is asked to press a scroll-wheel button. There are 8 of these trials and they are used in fMRI QA to determine alertness during the task. Each run last approximately 10 minutes.
The face stimuli used for version 3 (in place of NS1 or NS2 above depending on conditioning order)
Face actor 1077:
Face actor 1086:
There was then a subsequent reversal run (run2), which followed the same paradigm as run1 except the previously neutrally-conditioned stimulus (CS-) becomes paired with the scream with 50% contingency making it the run2 CS+. Inversely, the aversively-conditioned stimulus (CS+) from run1 is no longer paired with the scream, making it the run2 CS- stimulus.
There were four differing orders or versions of the design3 task which will be described below. Note this wiki focuses on the design3 or version3 fMRI task. There were two previous fMRI designs used earlier in the study (version1 and version2) that will also be described in more detail below.
Task Design Order:
The fear conditioning paradigm had four different orders (identified by the "Pairing number" and "event array") which each participant was assigned to in a counter-balanced manner prior to the scan. This order determined which of the face stimuli (actor face 1077 or actor face 1086) would be aversively conditioned and also determined if the aversive or neutrally conditioned stimulus would appear first in the task.
The first block, called ‘run1’, and the second block, called ‘reversal’ each have two immutable arrays associated with them. The first array, ‘image type,‘ consists of 1s and 2s (a different string for run1 and reversal) which is passed to the task presentation to determine the order of the stimuli presentation and the "aversive-ness" of the stimuli. The “pairing” number (found in [scanid]_Pairing.txt) determines which face image and sound condition to associate with these two numbers. The following table gives the mapping between "Pairing number" and the variable 'image-type' array characteristics:
- Pairing 0:
- imagetype = 1 corresponds to face 1077 and aversive tone
- imagetype = 2 corresponds to face 1086 and no tone
- Pairing 1:
- imagetype = 1 corresponds to face 1077 and no tone
- imagetype = 2 corresponds to face 1086 and aversive tone
- Pairing 2:
- imagetype = 1 corresponds to face 1086 and aversive tone
- imagetype = 2 corresponds to face 1077 and no tone
- Pairing 3:
- imagetype = 1 corresponds to face 1086 and no tone
- imagetype = 2 corresponds to face 1077 and aversive tone
This can also be gleaned from the [scanid]_run1_array.txt files (which displays the "event array"). The first two digits in the "event array" refer to which actor is the first face vs the second face; and the last two digits in the array refer to the task structure (ordering of aversive vs. neutral outcomes):
- 1 2 0 1
- 1 2 1 0
- 2 1 0 1
- 2 1 1 0
Task orders 1 and 3 have the same sound assignments (which affects "task structure")(as do task orders 0 and 2) but 1 and 3 have different actor-sound pairings and 0 and 2 have different actor pairings. Because of this, the fMRI analysis lumps across 0_2 and 1_3 because fMRI so far doesn't care about which actor it is but has to care about sound timing since that is what is modeled in first level. The aversive stimulus is presented at the same time in each of those order pairs (i.e. in pairing 1 and 3 the neutral stimulus occurs first). Therefore, in fMRI analyses, order is defined as either order 0_2 (which consists of participants that have either pairing 0 or pairing 2) or order 1_3 (which consists of participants that have either pairing 1 or pairing 3).
So the mapping from "event array" to "Pairing number" to "fMRI order" is as follows:
- 1 2 0 1 -> Pairing 0 -> Order 0_2
- 1 2 1 0 -> Pairing 1 -> Order 1_3
- 2 1 0 1 -> Pairing 2 -> Order 0_2
- 2 1 1 0 -> Pairing 3 -> Order 1_3
Task Stimulus Timing:
For each order, the files of stimulus times (stick files) are saved in /data/joy/BBL/studies/conte/fmriDesignFiles in directories for both orders (order0_2 and order1_3). Each face stimulus file is named with the face presentation order (1 or 2), the condition of the face (aversive or notone/neutral), and the model that the timing uses (i.e. mini block, split half, etc.). For example, face1_aversive_mini_block.txt corresponds to order0_2 because the first face to be presented (face1) is the aversive face (aversive) and the timing is used to model a mini block design. The first column denotes the stimulus onset times, the second column is the length of time the stimulus is present on the screen and the third column is a timing model column (this will be 1 except in linear timing models). See the face1_aversive_mini_block.txt stick file for order0_2 below.
6 6 1
14 6 1
90 6 1
122 6 1
228 6 1
244 6 1
276 6 1
292 6 1
334 6 1
364 6 1
448 6 1
482 6 1
There are also timing files for the catch trials, the aversive scream tone itself, all face stimuli (any_stim, which is the same for both orders), and the flashing screen that occurs at the end of the task.
Note run1 and run2 reversal tasks have different timings. So any_stim timings are same for order 1_3 run1 and order 0_2 run1, but different from order 1_3 run2 reversal/order 0_2 run2 reversal. e.g, the first face in run1 occurs at 6 seconds (24 seconds after first trigger), but in run2 reversal it occurs at 2 seconds (20 seconds after first trigger)- this means that reversal effects could theoretically be different from run1 effects on this basis.
See each stick file for timing of stimuli.
Task Stimulus Order:
Below is a table that details the ordering of the face stimuli for each order for all 48 trials:
Trial Order 0_2 Order 1_3
1 CS+p CS-
2 CS+p CS-
3 CS- CS+p
4 CS- CS+p
5 CS+u CS-
6 CS+u CS-
7 CS- CS+u
8 CS- CS+u
9 CS- CS+p
10 CS+p CS-
Catch Trial
11 CS- CS+p
Catch Trial
12 CS+p CS-
13 CS- CS+p
14 CS- CS+u
15 CS+u CS-
16 CS- CS+u
17 CS+u CS-
18 CS- CS+u
19 CS+u CS-
Catch Trial
20 CS+u CS-
21 CS+u CS-
22 CS+p CS-
23 CS- CS+p
24 CS+p CS-
Catch Trial
25 CS- CS+u
26 CS- CS+u
27 CS+p CS-
28 CS+u CS-
29 CS+p CS-
30 CS- CS+u
31 CS- CS+p
32 CS- CS+u
Catch Trial
33 CS+p CS-
34 CS- CS+p
35 CS+u CS-
36 CS+p CS-
37 CS+u CS-
Catch Trial
38 CS- CS+p
39 CS+u CS-
40 CS- CS+u
41 CS- CS+p
42 CS- CS+u
Catch Trial
43 CS- CS+u
44 CS+p CS-
45 CS- CS+p
Catch Trial
46 CS+u CS-
47 CS+p CS-
48 CS- CS+p
Logfiles:
All the information you need about a participant's fMRI fear conditioning run can be gained from the logfile that is output automatically by the presentation code. The output for run1 fear conditioning is in [scanid]-fearConditioning_run1_wheel.log and in [scanid]-fearConditioning_rev_wheel.log for run2 reversal. Both of these files are saved in xnat under the fear conditioning run associated files portion, and are downloaded with the download script.
Note: the naming convention of the fear conditioning logfiles will sometimes be incorrect (i.e. they may mention "Pairing0" in the name when the participant is not pairing 0). The safest way to determine order is by looking in the log file itself for the first face stimulus (face1077 or face1086 and pairing aversive or notone).
If the variable 'imagetype' described above were the only array used to define what is displayed at each stimulus event, we would not be able to play the aversive tone only 50% of the time the aversive face is displayed. In order to control the presentation of the tone systematically, another array, “event” is created. "Event" controls the type of trial is displayed to the participant. The event array consists of the numbers 0,1,2,3,4, each with a unique meaning.
- "0" corresponds to a crosshair trial.
- “1” corresponds to a trial where the stimulus is muted
- i.e., if the current imageType is 2, and we are in pairing 1, no tone will result, even though imageType = 2 is associated with an aversive tone.
- “2” corresponds to a trial where the corresponding imageType is displayed as is, with all its associated characteristics.
- “3” corresponds to a ‘catch’ trial, where the cross hair turns green and the user is prompted to press the response button.
- “4” corresponds to a ‘buffer’ trial, added for convenience, that allows event types 1 and 2 to be lengthened to multiple 2 second chunks.
Below is an example logfile for a pairing 0 run1 fear conditioning run. Logfiles are very long so I have detailed the important lines by annotating them below. I have also deleted the bulk of the middle of the logfile to save space.
cat /data/joy/BBL/studies/conte/rawData/94848/*x10178/associated_files/ratings/post/10178-fearConditioning_run1_wheel_Pairing0.log
Scenario - fearConditioning_run1_wheel_Pairing0
Logfile written -
Subject Trial Event Type Code Time TTime Uncertainty Duration Uncertainty ReqTime ReqDur Stim Type Pair Index
10178 0 Manual first_trigger 216978 216978
10178 0 Pulse 1 216978 216978 1
10178 0 Manual pulseCount_1 217336 217336
10178 1 Picture face_array_1077_1086 217479 0 1 30164 2 0 next other 0
10178 1 Pulse 1 246977 29498 1
10178 2 Pulse 1 276976 29333 1
10178 3 Pulse 1 306975 29168 1
10178 3 Manual flashing_screen 307807 30000
10178 4 Pulse 1 336974 29004 1
10178 4 Manual press_button_instructions 337970 30000
10178 5 Pulse 1 366973 28839 1
10178 5 Manual pulseCount_6 368134 30000
10178 5 Pulse 1 396972 58838 1
10178 5 Manual pulseCount_7 396973 58839
10178 5 Manual cross 396973 58839
10178 6 Manual cross 416980 19852
10178 7 Manual pulseCount_8 426972 9845
10178 7 Pulse 1 426972 9845 1
10178 7 Manual cross 436990 19863
10178 8 Manual pulseCount_9 456971 19846
10178 8 Manual Trial_1 456971 19846
10178 8 Pulse 1 456971 19846 1
10178 8 Manual what_face_1077_tone 456972 19847
The above line tells you the face that will be presented next and if there will be a scream with it or not (in this case the first face is 1077 and it is aversive)
10178 9 Picture face_on 457123 0 1 20164 3 0 next other 0
10178 9 Response 2 457371 248 1
10178 9 Manual Response2 477123 20000
10178 10 Picture face_on 477287 0 2 20165 3 0 next other 0
10178 10 Pulse 1 486970 9683 1
10178 10 Manual pulseCount_10 497287 20000
10178 11 Picture face_on 497452 0 1 20165 2 0 next other 0
10178 11 Sound aversive_tone_on 507452 10000 2 10000 other 0
The above line is where the aversive sound is played
10178 11 Pulse 1 516969 19517 1
10178 11 Manual pulseCount_11 517543 20091
10178 11 Manual cross 517544 20092
10178 12 Manual Trial_2 537550 19933
10178 12 Manual what_face_1077_tone 537550 19933
10178 13 Picture face_on 537615 0 1 20165 2 0 next other 0
10178 13 Pulse 1 546969 9354 1
10178 13 Manual pulseCount_12 557615 20000
10178 14 Picture face_on 557780 0 1 20164 2 0 next other 0
10178 14 Pulse 1 576967 19187 1
10178 14 Manual pulseCount_13 577780 20000
10178 15 Picture face_on 577944 0 1 20165 2 0 next other 0
10178 15 Sound aversive_tone_on 587944 10000 2 10000 other 0
10178 15 Manual cross 598040 20096
............ I have deleted the middle here for space ....................
10178 50 Pulse 1 1356956 19585 1
10178 50 Manual pulseCount_39 1357418 20047
10178 50 Manual cross 1357418 20047
10178 51 Manual cross 1377420 19884
10178 52 Manual pulseCount_40 1386955 9421
10178 52 Pulse 1 1386955 9421 1
10178 53 Manual green_cross 1397699 190
The above line denotes a catch trial
10178 53 Manual Response2 1408035 10526
The above line records if the participant responded (pressed the scroll-wheel as instructed) to the catch trial
10178 53 Response 2 1408035 10526 1
10178 54 Manual pulseCount_41 1416954 8757
............ I have deleted the middle here for space ....................
10178 249 Manual events_complete 5337652 20000
10178 250 Pulse 1 5346866 9049 1
10178 251 Pulse 1 5376865 8884 1
10178 251 Manual flashing_screen 5397981 30000
The above line is the flashing screen at end of the task
10178 252 Pulse 1 5406864 8719 1
10178 252 Pulse 1 5436863 38718 1
10178 252 Pulse 1 5466863 68718 1
10178 252 Manual End_of_task 5478145 80000
QA procedures:
Motion:
Motion was extracted by running FSL feat and examining the "prefiltered_func_data_mcf_rel_mean.rms" file in their feat output. If motion was greater than 0.3, the participant was excluded. Six total subjects (1 NC, 2 CR, 2 P) were excluded based on run1 high motion.
Sleepiness:
The participant's alertness during the task was also considered by reading any notes by the scan acquisition team in Oracle regarding if the participant fell asleep or not. Only one participant (CR group) was excluded based on falling asleep.
Major Differences between Versions:
There are several differences between version1, version2, and version3 of the fear conditioning task. Detailed below are the major differences in the fear conditioning task between versions. There are more minor timing differences as well, but you should examine the logfiles themselves or presentation code if you would like to learn more about the exact stimulus timing differences. Note: differences in ratings and other aspects of the study are recorded in their respective wikis.
- Version 1 ran from 12/20/2012 to 03/16/2013 and scanids 7563 to 7831; 22 fMRI subjects were collected
- Version 2 ran from 6/20/2013 to 03/10/2014 and scanids 8150 to 8779; 41 fMRI subjects were collected
- Version 3 ran from 7/12/2014 to the end of the study (May 2017) and scanid 8964 to 10564; 120 fMRI subjects were collected
Version1 to Version2:
-
Version 1 had a maintenance run, so that there were three fear conditioning tasks: conditioning (run1 V3R1), maintenance (run2 V3R2), and reversal (run3 V3R3). The maintenance run was identical to the conditioning run, and served as a conditioning reinforcement. For version 2, the maintenance run was removed so that there were only conditioning (run1 V3R1) and reversal (run2 reversal V3R2) runs. This was done because many subjects were too sleepy and we wanted to shorten the scan time to 1 hour.
-
Subjective post scan ratings of the stimuli (tone, scream, and sleepiness) were also recorded in version 2 but not version 1.
-
The tone was also likely tweaked (increased pitch, lowered amplitude), so that the neutral tone was potentially less unpleasant. The scream was changed from a white-noise + scream to a mix of male and female screams or "double scream". (Although it's possible this change was done from pilot to version1 and not from version1 to version2).
Version2 to Version3:
-
The tone paired with the neutral stimuli in version 2 was removed in version 3. In version 3 therefore, no sound is paired with the aversive stimulus. This was done because participants were reporting the neutral tone approximately as unpleasant or loud as the aversive stimulus and therefore there was concern that the fear conditioning would not occur appropriately because both conditions were aversive.
-
The task was simplified from utilizing 4 face stimuli (actor faces 1023, 1057, 1077, and 1086) two aversive and two neutral to only 2 face stimuli (actor faces 1077 and 1086) one aversive and one neutral.
-
The acquisition volume was reduced from 205 volumes in version 2 to 176 volumes in version 3.
-
The task had 96 trials in version 2 and 48 trials in verion 3 (half of previously because there were half the number of faces presented).
-
More advanced MRI earphones were purchased to both to maintain high-fidelity sound quality/volume transmission and to suppress the noise from the scanner.
-
The duration between the face CS onset and scream US onset was increased from 1 second to 5 seconds, to separate the face onset and scream more and also to make the task more similar to the EEG task. This also caused the trial duration itself to increase from 2 seconds to 6 seconds in length. In versions 1 and 2 the face was on for 2 seconds total, for 1 second alone then co-terminating with the tone for 1 second.
-
It is also probable that counterbalancing of face stimulus order and actor assignment was done, yielding 4 task versions (pairings 0,1,2,3) with 2 aversive/neutral sequences, in a 2x2 with the 2 faces.
-
Catch trials were reduced from 10 to 8.
Pilot study:
Note there was also a CONTE pilot study. The CONTE grant was submitted Sep 2011. The grant reports pilot data from 8 healthy controls. The pilot study included a conditioning phase (v1vr1), maintenance (basically identical to conditioning, maybe a different specific trial sequence) (v1vr2), and a reversal phase (v1vr2_reversal). Per DW memory there were both face and object neutral CSs in pilots, the object CSs were dropped from plan after initial analyses did not show amygdala average-task-activation during object condition, and grant proposal only describes face CS. Pilot data on monstrum is in /import/monstrum/conte2_pilot/subjects and the subdirectory patients, and includes 9 presumably healthy controls and 5 patients, Scanids span 05055 through 05364.