The Real-time fMRI Neurofeedback Based Stratification of Default Network Regulation Neuroimaging Data Repository

Amalia McDonald, Jordan Muraskin, Nicholas T. Van Dam, Caroline Froehlich, Benjamin Puccio, John Pellman, Clemens Bauer, Alexis Akeyson, Melissa M. Breland, Vince D. Calhoun, Stephen Carter, Tiffany Chang, Chelsea Gessner, Alyssa Gianonne, Steven Giavasis, Jamie Glass, Steven Homan, Margaret King, Melissa Kramer, Drew Landis, Alexis Lieval, Jonathan Lisinski, Anna MacKay‐Brandt, Brittny Miller, Laura J. Panek, Hayley Reed, Christine Santiago, Eszter Schoell, Richard Sinnig, Melissa Sital, Elise C. Taverna, Russell H. Tobe, Kristin Dietz Trautman, Betty Varghese, Lauren Walden, Runtang Wang, Abigail B. Waters, Dylan Wood, F. Xavier Castellanos, Bennett Leventhal, Stanley J. Colcombe, Stephen M. LaConte, Michael P. Milham, R Cameron CraddocbioRxiv 2016 · DOI

TL;DR

This paper introduces a new, openly shared repository of real-time fMRI neurofeedback data designed to study individual differences in the brain's "default mode network" (DMN). The repository includes diverse fMRI tasks and extensive behavioral/cognitive measures, offering a valuable resource for understanding DMN regulation and its implications for various conditions.

What Problem Does This Paper Address?

The default mode network (DMN) is a set of brain regions active during rest, implicated in spontaneous thought and emotion regulation. While alterations in the DMN are linked to neuropsychiatric conditions, most studies focus on its resting state or deactivation during tasks. This approach fails to distinguish between an individual's capacity to modulate the DMN and their tendency to do so. Understanding these individual differences is crucial, as they may have distinct implications for mental health. Real-time fMRI (rt-fMRI) neurofeedback offers a powerful tool to directly assess and train DMN modulation, but requires robust datasets to explore these individual variations. This paper addresses this gap by releasing a comprehensive dataset collected using rt-fMRI neurofeedback and related tasks, alongside rich phenotypic data.

Methods

The study employed a multi-faceted approach combining various fMRI tasks and extensive behavioral/cognitive assessments. Key methods include:

  • Functional MRI (fMRI) Acquisition: Standardized fMRI scans were acquired using a 3T Siemens scanner, including anatomical (MPRAGE) and functional (EPI) sequences.
  • Task-Based fMRI:
    • Multi-Source Interference Task (MSIT): Used to assess DMN deactivation.
  • Moral Dilemma Task (MD): Used to assess DMN activation.
  • Resting State fMRI: Collected to establish baseline DMN activity.
  • Real-Time fMRI (rt-fMRI) Neurofeedback: A novel paradigm was developed to provide participants with real-time visual feedback on their DMN activity, allowing them to attempt to increase or decrease it. This involved a classifier-based approach trained on resting-state data to extract DMN activity TR-by-TR.
  • Data Preprocessing: A standardized pipeline (C-PAC) was used for preprocessing, including skull stripping, segmentation, normalization, and motion correction.
  • Behavioral and Cognitive Assessments: A battery of self-report measures (e.g., Affect Intensity Measure, Emotion Regulation Questionnaire, Penn State Worry Questionnaire) and behavioral tasks (e.g., Rapid Visual Information Processing - RVIP) were administered to capture individual differences in emotion regulation, mind-wandering, and cognitive function.
  • Quality Assessment: Standardized metrics (e.g., SNR, CNR, FBER, DVARS, RMSD) were used to evaluate the spatial and temporal quality of the fMRI data, compared against established benchmarks.

Datasets

The paper introduces the Default Network Regulation Neuroimaging Repository, which contains:

  • Functional MRI Data:
    • Resting-state fMRI scans.
  • fMRI data from the Multi-Source Interference Task (MSIT).
  • fMRI data from the Moral Dilemma Task (MD).
  • Real-time fMRI neurofeedback data.
  • Anatomical MRI scans.
  • Behavioral and Cognitive Data:
    • Self-report measures related to perseverative thinking, emotion regulation, and imaginative processes.
  • Behavioral measures, including the Rapid Visual Information Processing (RVIP) task.
  • Demographic and Clinical Data: Basic phenotypic data (age, sex, handedness) and diagnostic information (SCID-IV).

This repository is harmonized with and distributed alongside the Enhanced Nathan Kline Institute-Rockland Sample (NKI-RS). The data is available through the COINS Data Exchange, NITRC, and Amazon Web Services. The paper reports on technical validation from n=125 participants of a final targeted sample of 180.

Key Insights and Ideas

  • DMN Modulation is Highly Variable: Group-level analysis of the neurofeedback data revealed significant inter-subject variability in the ability to modulate DMN activity, highlighting the importance of studying individual differences.
  • Sleep Significantly Impacts DMN Modulation: Participants who reported sleep during resting-state or neurofeedback scans showed lower correlations with the ideal task model, suggesting that sleep can impair DMN regulation. Removing these participants improved the correlation, underscoring the need to account for sleep in analyses.
  • Task-Specific DMN Activation/Deactivation Confirmed: Preliminary analyses confirmed that the MSIT task reliably deactivated the DMN, while the Moral Dilemma task reliably activated it, validating the experimental design.
  • Data Quality is Generally Good, but Compliance is a Concern: While standard quality metrics indicated good data quality, a significant portion of participants (up to 73 out of 125) may need to be excluded from analyses due to poor task performance, sleep, or technical issues. This highlights a critical challenge in neuroimaging research.
  • Repository as a Resource: The primary contribution is the creation and release of this comprehensive, multi-modal dataset, enabling researchers to investigate DMN function, regulation, and its relationship to various cognitive and clinical factors.

Key Results

  • Neurofeedback Effectiveness: Group-level analysis showed participants could modulate their DMN, with a higher correlation to the ideal task model when participants who reported sleep were removed (R²=0.68 vs. R²=0.36).
  • Task Validation: The Moral Dilemma task showed DMN activation, and the MSIT task showed DMN deactivation, as predicted.
  • Behavioral Performance: RVIP and MSIT performance was generally consistent with previous literature, though slightly worse in this cohort, potentially due to the inclusion of participants with psychiatric diagnoses.
  • Data Loss: Approximately 50 participants had unusable data due to poor performance, sleep, or technical issues. Sleep alone accounted for the exclusion of 46 participants.

Limitations and Future Work

  • Participant Compliance: The high rate of data loss due to poor performance, sleep, and technical issues is a significant limitation. The authors suggest that using younger, healthier, or more experienced participants might improve compliance. They also emphasize the need for better monitoring of wakefulness during scans and potentially using objective sleep measures.
  • Sleep Impact: While the study highlights the impact of sleep, the repository allows researchers to choose whether to include or exclude participants based on self-reported sleep, offering flexibility for different research questions.
  • Future Directions: The authors encourage users to provide feedback for improving the resource and are interested in learning about the research conducted with the data. The repository is intended to be updated regularly with new data. The data is expected to be particularly useful for students and junior researchers for hypothesis testing, learning new techniques, and as pilot data for grant applications.

Summary updated

Abstract

Abstract This data descriptor describes a repository of openly shared data from an experiment to assess inter-individual differences in default mode network (DMN) activity. This repository includes cross-sectional functional magnetic resonance imaging (fMRI) data from the Multi Source Interference Task, to assess DMN deactivation, the Moral Dilemma Task, to assess DMN activation, a resting state fMRI scan, and a DMN neurofeedback paradigm, to assess DMN modulation, along with accompanying behavioral and cognitive measures. We report technical validation from n=125 participants of the final targeted sample of 180 participants. Each session includes acquisition of one whole-brain anatomical scan and whole-brain echo-planar imaging (EPI) scans, acquired during the aforementioned tasks and resting state. The data includes several self-report measures related to perseverative thinking, emotion regulation, and imaginative processes, along with a behavioral measure of rapid visual information processing. Technical validation of the data confirms that the tasks deactivate and activate the DMN as expected. Group level analysis of the neurofeedback data indicates that the participants are able to modulate their DMN with considerable inter-subject variability. Preliminary analysis of behavioral responses and specifically self-reported sleep indicate that as many as 73 participants may need to be excluded from an analysis depending on the hypothesis being tested.
BibTeX
@misc{mcdonald2016real,
  author = {Amalia McDonald and Jordan Muraskin and Nicholas T. Van Dam and Caroline Froehlich and Benjamin Puccio and John Pellman and Clemens Bauer and Alexis Akeyson and Melissa M. Breland and Vince D. Calhoun and Stephen Carter and Tiffany Chang and Chelsea Gessner and Alyssa Gianonne and Steven Giavasis and Jamie Glass and Steven Homan and Margaret King and Melissa Kramer and Drew Landis and Alexis Lieval and Jonathan Lisinski and Anna MacKay‐Brandt and Brittny Miller and Laura J. Panek and Hayley Reed and Christine Santiago and Eszter Schoell and Richard Sinnig and Melissa Sital and Elise C. Taverna and Russell H. Tobe and Kristin Dietz Trautman and Betty Varghese and Lauren Walden and Runtang Wang and Abigail B. Waters and Dylan Wood and F. Xavier Castellanos and Bennett Leventhal and Stanley J. Colcombe and Stephen M. LaConte and Michael P. Milham and R Cameron Craddoc},
  title = {{The Real-time fMRI Neurofeedback Based Stratification of Default Network Regulation Neuroimaging Data Repository}},
  year = {2016},
  doi = {10.1101/075275},
  howpublished = {bioRxiv},
}
Figure