This page includes supplemental materials for the paper, FBIRN Recommendations for Multi-Center fMRI Studies:

In recognition of concerns identified during the development and pretesting of the stimulus/behavioral methodology, the FBIRN project modified its behavioral software and many of its on-site procedures part way through the study.  The paradigm methods we implemented to facilitate multi-center data collection and improve overall data quality were:

1)    Flexible software — We adopted the CIGAL program (Voyvodic, 1999; http://wiki.biac.duke.edu/wiki/doku.php/jvs:cigal) as our standard behavioral task software at all collection sites.  This change was primarily motivated by CIGAL’s flexibility to accommodate different task designs, while simultaneously providing accurate recordings of multiple different analog and digital input signals and visualizing the performance data in real time during scanning. This allowed immediate remedying of performance problems such as subject inattention or using the wrong buttons. When installing the software, all sites were able to configure it to recognize their particular button box, scanner trigger, video, and audio equipment.

2)    Scanner timing – We simplified control parameters to allow each site to specify how many scanner trigger pulses should be ignored so that the start of the task would be synchronized with the acquisition of the first saved MR image. Despite this, errors were occasionally made resulting in confusion about task and scanner timing.  The solution we added was to have the paradigm software record scanner trigger pulses throughout the task, and a couple of seconds past the last expected scanner pulse. If scanner pulses did not stop exactly when expected, the paradigm software presents an alert indicating there is a timing problem. By recording both the start and end of the scanner imaging series, along with the scanner’s own data records of how many images were acquired, has effectively removed uncertainty about task timing at all sites.

3)    Response accuracy – When the FBIRN tasks are run, a simple initialization routine enables the user to calibrate which particular response button will be used for each type of user response, and to verify that video and auditory settings are appropriate. This approach allows the software to be flexible enough to deal with diverse hardware configurations, while still imposing standardization on the procedures performed at run time. It also ensures that different input signals are mapped to standardized output values in behavioral log files. Although the accuracy of response timing remains partially dependent on the mechanics and electronics of the particular response device and its interface, the CIGAL software itself records all incoming data with sub-millisecond accuracy.

4)    Physiological monitoring – FBIRN implemented routine recording of both respiratory and cardiac physiological oscillations throughout MR scanning by making use of CIGAL’s ability to automatically record multiple input signals simultaneously. For this, a simple inexpensive USB-based analog-to-digital converter device (Model USB-1208FS, Measurement Computing, www.mccdaq.com) was installed at all sites along with an FBIRN-provided simple cable connector box. Each site was also equipped with respiration belt and finger-cuff pulse-oximeter transducers and a cardiac signal amplifier (Biopac, www.biopac.com) to provide a robust standardized physiological interface. Although the installation of this equipment involved some coordinated effort to standardize suitable grounding, once installed it has been used successfully at every site simply by connecting the transducers to each subject. CIGAL records the input signals automatically with accurate synchronization to both task and MR image acquisition.

5)    Data quality control – To improve quality control (QC) we implemented several data checking and documentation features into our paradigm software. The first involved real-time checking for simple errors in data entry, such as requiring subject IDs to be entered twice, forcing operator-supplied subject IDs and scan series labels to match expected guidelines, and providing alert messages if trigger timing, subject responses, or data labels were significantly outside expected ranges. The second involved using CIGAL’s real-time features to present stimuli to the subject via one video screen and simultaneously using a second video monitor to provide a continuous oscilloscope-like display of the subject’s task performance, physiological status, and scanner trigger pulses throughout every fMRI scan. Where available, subject eye position is also shown on the data feedback display. The ability to continuously monitor all data as they are being collected greatly increases the likelihood that data acquisition problems will be detected and fixed when they occur. The third QC feature that has proven very useful is that our paradigm software provides a comprehensive log of each behavioral session.  This involves recording not only the specific response and physiological data for the task itself, but also an accurate log of every individual stimulus event, every button press (whether expected or not), all software and hardware parameter settings, and every key press or mouse click performed by the computer operator for the entire session. Having all this information makes it possible to make corrections during scanning as well as later reconstruct much of what happened during every scan at every site.

6)    Distribution and version control – Because many problems in a multi-site study are due to slightly different versions of software being used at different sites, we improved our software version control procedures to facilitate version synchronization across sites and to ensure that all behavioral data collected are tagged with version identifiers for all software components. For the FBIRN, this was accomplished by implementing a customized turnkey program that automatically compares all local software files against a software repository at a central web site. Any missing, modified, or out of date files are flagged and can be quickly updated if necessary. All software changes are logged and a site can easily choose to go back to any previous version if desired. This simple interactive web interface has greatly simplified multi-center development, distribution, and maintenance of all our behavioral software.

Overall, the lesson for multi-center task paradigms is: standardize as much as possible, include multiple real-time procedures for monitoring data quality, and accurately record as much of what happens during each session as possible.

This page includes supplemental materials for the paper, FBIRN Recommendations for Multi-Center fMRI Studies:

In recognition of concerns identified during the development and pretesting of the stimulus/behavioral methodology, the FBIRN project modified its behavioral software and many of its on-site procedures part way through the study.  The paradigm methods we implemented to facilitate multi-center data collection and improve overall data quality were:

1)    Flexible software — We adopted the CIGAL program (Voyvodic, 1999; http://wiki.biac.duke.edu/wiki/doku.php/jvs:cigal) as our standard behavioral task software at all collection sites.  This change was primarily motivated by CIGAL’s flexibility to accommodate different task designs, while simultaneously providing accurate recordings of multiple different analog and digital input signals and visualizing the performance data in real time during scanning. This allowed immediate remedying of performance problems such as subject inattention or using the wrong buttons. When installing the software, all sites were able to configure it to recognize their particular button box, scanner trigger, video, and audio equipment.

2)    Scanner timing – We simplified control parameters to allow each site to specify how many scanner trigger pulses should be ignored so that the start of the task would be synchronized with the acquisition of the first saved MR image. Despite this, errors were occasionally made resulting in confusion about task and scanner timing.  The solution we added was to have the paradigm software record scanner trigger pulses throughout the task, and a couple of seconds past the last expected scanner pulse. If scanner pulses did not stop exactly when expected, the paradigm software presents an alert indicating there is a timing problem. By recording both the start and end of the scanner imaging series, along with the scanner’s own data records of how many images were acquired, has effectively removed uncertainty about task timing at all sites.

3)    Response accuracy – When the FBIRN tasks are run, a simple initialization routine enables the user to calibrate which particular response button will be used for each type of user response, and to verify that video and auditory settings are appropriate. This approach allows the software to be flexible enough to deal with diverse hardware configurations, while still imposing standardization on the procedures performed at run time. It also ensures that different input signals are mapped to standardized output values in behavioral log files. Although the accuracy of response timing remains partially dependent on the mechanics and electronics of the particular response device and its interface, the CIGAL software itself records all incoming data with sub-millisecond accuracy.

4)    Physiological monitoring – FBIRN implemented routine recording of both respiratory and cardiac physiological oscillations throughout MR scanning by making use of CIGAL’s ability to automatically record multiple input signals simultaneously. For this, a simple inexpensive USB-based analog-to-digital converter device (Model USB-1208FS, Measurement Computing, www.mccdaq.com) was installed at all sites along with an FBIRN-provided simple cable connector box. Each site was also equipped with respiration belt and finger-cuff pulse-oximeter transducers and a cardiac signal amplifier (Biopac, www.biopac.com) to provide a robust standardized physiological interface. Although the installation of this equipment involved some coordinated effort to standardize suitable grounding, once installed it has been used successfully at every site simply by connecting the transducers to each subject. CIGAL records the input signals automatically with accurate synchronization to both task and MR image acquisition.

5)    Data quality control – To improve quality control (QC) we implemented several data checking and documentation features into our paradigm software. The first involved real-time checking for simple errors in data entry, such as requiring subject IDs to be entered twice, forcing operator-supplied subject IDs and scan series labels to match expected guidelines, and providing alert messages if trigger timing, subject responses, or data labels were significantly outside expected ranges. The second involved using CIGAL’s real-time features to present stimuli to the subject via one video screen and simultaneously using a second video monitor to provide a continuous oscilloscope-like display of the subject’s task performance, physiological status, and scanner trigger pulses throughout every fMRI scan. Where available, subject eye position is also shown on the data feedback display. The ability to continuously monitor all data as they are being collected greatly increases the likelihood that data acquisition problems will be detected and fixed when they occur. The third QC feature that has proven very useful is that our paradigm software provides a comprehensive log of each behavioral session.  This involves recording not only the specific response and physiological data for the task itself, but also an accurate log of every individual stimulus event, every button press (whether expected or not), all software and hardware parameter settings, and every key press or mouse click performed by the computer operator for the entire session. Having all this information makes it possible to make corrections during scanning as well as later reconstruct much of what happened during every scan at every site.

6)    Distribution and version control – Because many problems in a multi-site study are due to slightly different versions of software being used at different sites, we improved our software version control procedures to facilitate version synchronization across sites and to ensure that all behavioral data collected are tagged with version identifiers for all software components. For the FBIRN, this was accomplished by implementing a customized turnkey program that automatically compares all local software files against a software repository at a central web site. Any missing, modified, or out of date files are flagged and can be quickly updated if necessary. All software changes are logged and a site can easily choose to go back to any previous version if desired. This simple interactive web interface has greatly simplified multi-center development, distribution, and maintenance of all our behavioral software.

Overall, the lesson for multi-center task paradigms is: standardize as much as possible, include multiple real-time procedures for monitoring data quality, and accurately record as much of what happens during each session as possible.

This page includes supplemental materials for the paper, FBIRN Recommendations for Multi-Center fMRI Studies:

In recognition of concerns identified during the development and pretesting of the stimulus/behavioral methodology, the FBIRN project modified its behavioral software and many of its on-site procedures part way through the study.  The paradigm methods we implemented to facilitate multi-center data collection and improve overall data quality were:

1)    Flexible software — We adopted the CIGAL program (Voyvodic, 1999; http://wiki.biac.duke.edu/wiki/doku.php/jvs:cigal) as our standard behavioral task software at all collection sites.  This change was primarily motivated by CIGAL’s flexibility to accommodate different task designs, while simultaneously providing accurate recordings of multiple different analog and digital input signals and visualizing the performance data in real time during scanning. This allowed immediate remedying of performance problems such as subject inattention or using the wrong buttons. When installing the software, all sites were able to configure it to recognize their particular button box, scanner trigger, video, and audio equipment.

2)    Scanner timing – We simplified control parameters to allow each site to specify how many scanner trigger pulses should be ignored so that the start of the task would be synchronized with the acquisition of the first saved MR image. Despite this, errors were occasionally made resulting in confusion about task and scanner timing.  The solution we added was to have the paradigm software record scanner trigger pulses throughout the task, and a couple of seconds past the last expected scanner pulse. If scanner pulses did not stop exactly when expected, the paradigm software presents an alert indicating there is a timing problem. By recording both the start and end of the scanner imaging series, along with the scanner’s own data records of how many images were acquired, has effectively removed uncertainty about task timing at all sites.

3)    Response accuracy – When the FBIRN tasks are run, a simple initialization routine enables the user to calibrate which particular response button will be used for each type of user response, and to verify that video and auditory settings are appropriate. This approach allows the software to be flexible enough to deal with diverse hardware configurations, while still imposing standardization on the procedures performed at run time. It also ensures that different input signals are mapped to standardized output values in behavioral log files. Although the accuracy of response timing remains partially dependent on the mechanics and electronics of the particular response device and its interface, the CIGAL software itself records all incoming data with sub-millisecond accuracy.

4)    Physiological monitoring – FBIRN implemented routine recording of both respiratory and cardiac physiological oscillations throughout MR scanning by making use of CIGAL’s ability to automatically record multiple input signals simultaneously. For this, a simple inexpensive USB-based analog-to-digital converter device (Model USB-1208FS, Measurement Computing, www.mccdaq.com) was installed at all sites along with an FBIRN-provided simple cable connector box. Each site was also equipped with respiration belt and finger-cuff pulse-oximeter transducers and a cardiac signal amplifier (Biopac, www.biopac.com) to provide a robust standardized physiological interface. Although the installation of this equipment involved some coordinated effort to standardize suitable grounding, once installed it has been used successfully at every site simply by connecting the transducers to each subject. CIGAL records the input signals automatically with accurate synchronization to both task and MR image acquisition.

5)    Data quality control – To improve quality control (QC) we implemented several data checking and documentation features into our paradigm software. The first involved real-time checking for simple errors in data entry, such as requiring subject IDs to be entered twice, forcing operator-supplied subject IDs and scan series labels to match expected guidelines, and providing alert messages if trigger timing, subject responses, or data labels were significantly outside expected ranges. The second involved using CIGAL’s real-time features to present stimuli to the subject via one video screen and simultaneously using a second video monitor to provide a continuous oscilloscope-like display of the subject’s task performance, physiological status, and scanner trigger pulses throughout every fMRI scan. Where available, subject eye position is also shown on the data feedback display. The ability to continuously monitor all data as they are being collected greatly increases the likelihood that data acquisition problems will be detected and fixed when they occur. The third QC feature that has proven very useful is that our paradigm software provides a comprehensive log of each behavioral session.  This involves recording not only the specific response and physiological data for the task itself, but also an accurate log of every individual stimulus event, every button press (whether expected or not), all software and hardware parameter settings, and every key press or mouse click performed by the computer operator for the entire session. Having all this information makes it possible to make corrections during scanning as well as later reconstruct much of what happened during every scan at every site.

6)    Distribution and version control – Because many problems in a multi-site study are due to slightly different versions of software being used at different sites, we improved our software version control procedures to facilitate version synchronization across sites and to ensure that all behavioral data collected are tagged with version identifiers for all software components. For the FBIRN, this was accomplished by implementing a customized turnkey program that automatically compares all local software files against a software repository at a central web site. Any missing, modified, or out of date files are flagged and can be quickly updated if necessary. All software changes are logged and a site can easily choose to go back to any previous version if desired. This simple interactive web interface has greatly simplified multi-center development, distribution, and maintenance of all our behavioral software.

Overall, the lesson for multi-center task paradigms is: standardize as much as possible, include multiple real-time procedures for monitoring data quality, and accurately record as much of what happens during each session as possible.