Contents

Overview

Overview

CompSAN is very excited to announce the inaugural SANS 2022 Naturalistic fMRI Data Analysis Challenge. Naturalistic functional magnetic resonance imaging (fMRI) provides a unique opportunity to study the neural basis of social, cognitive, and affective processing. Data from naturalistic conditions are rich and dynamic, and present both an opportunity and a challenge to analysis methods.

This data challenge provides the chance to demonstrate the application of new algorithms and models to naturalistic fMRI data to shed new light on social and affective neural processing at the intra- and/or inter-subject level. We welcome submissions from all levels of imaging experience, from individual trainees new to naturalistic data to groups well established in this approach.

The goals of the challenge include:

  1. Increasing a diversity of analysis approaches in SANS community

  2. Advancing theory development based on computational approaches

  3. Providing a forum for trainees of all levels to develop and explore new computational/analytic skills.

The top submissions will have an opportunity to present their project as an oral presentation during the Methods/Computational symposium at SANS 2022, to be held on May 5th, 2022 at 3.30-5pm US EDT/12.30-2pm US PDT (https://socialaffectiveneuro.org/conference/) and will share a $5,000 prize from the Consortium for Interacting Minds at Dartmouth College.

While we are framing this challenge as a competition, we hope that this will be a positive experience for everyone involved.

Judging Criteria

There is no objective benchmark or specific problem in the data to discover. Instead, we are primarily evaluating projects on creativity and utility.

Entries for the competitive portion of the challenge will be judged on:

  1. Novelty/originality of analysis approach (Has this been done before?)

  2. The connection between analysis approach and social and affective neuroscience theory/topics (What does this tell us about the brain or mind?)

  3. Robustness and replicability of results. (How robust are the results relative to the analytic choices? How much did you overfit your analysis to the dataset?)

The judging panel includes:

Winners and prizes will be announced during the computational symposium at the 2022 SANS conference. Entrants are also welcome to opt-out of the competition and enter just for fun/learning.

How to enter the challenge:

  1. First, download the FridayNightLights_Study2 Dataset dataset here:

  2. If you are new to naturalistic data processing, we encourage you to try applying methods described in the Naturalistic Data Analysis tutorials

  3. Form a team, or work solo. Once you are ready, register your team/individual entry. Registration is free, and is now open. You may enter as late as the day of submission.

  4. To be a participant in the competition, at least one member of your team will need to register for the main SANS 2022 Conference. In addition, you will need to register for the main conference to attend the computational symposium where winners, prizes, and more will be announced.

  5. Post questions to the CompSAN Discourse Page.

  6. Submit a poster and extended abstract (3 pages), and notebook/matlab script (w details on associated dependencies). The abstract should include separate sections for background, methods, results and conclusions. Entrants will have the option of making their notebook/scripts publicly available. The deadline for submissions has been extended to midnight (EST) on April 15th (Originally 4/2/22). At the time of submission, you will be given the opportunity to opt-out of competition.

  7. Especially important for trainees (at all stages): complete a survey on your learning experience and the challenges you faced.

Struggling for inspiration? Here are some possible ideas:

One exciting aspect of this challenge is that there are really an infinite number of possible things to try, which unfortunately can make it overwhelming to figure out where to begin. For inspiration for how different methods have been used to study naturalistic data more broadly, check out the Naturalistic Data Analysis Tutorials. You might be able to apply or modify one of those methods to ask a new question for social and affective neuroscience. Alternatively, you might think of a new questions and then try to create a new analysis method to answer your question.

Here are a few project brainstorms that might spark some ideas.

Encoding models of social cognition using naturalistic conditions

Encoding models have been of great value in understanding sensory and perceptual responses to naturalistic stimuli. As the sophistication of computational models of social cognition increases, so does the potential to apply these as encoding models to understand the neural basis of social cognition. However, unlike sensory systems social cognition is often not directly related to the stimuli itself, which will likely require more sophisticated models to predict how people are encoding social information.

Dynamics and latent states in naturalistic imaging data

Affective states and associated brain activity are highly dynamic. Spatial differences in the temporal dynamics of brain responses can lead to hidden or latent states not fully captured by standard, statically-defined patterns or networks. There are a number of methods that might help uncover these states. Is the state-space discrete or continuous? What level do you think this information is being represented in the brain? Within an ROI? Patterns across the brain? or maybe in network dynamics?

Nonlinear and/or time-warping methods for intersubject correlations

Standard methods for measuring intersubject synchrony typically rely on linear correlations in the time domain. However, individual differences in the dynamics of spontaneous activity might not be noise, and methods that can account for speed/timing differences and nonlinear relationships between subjects might reveal new information about group-level synchrony.

Questions

Please post any questions that arise on our Discourse Page. This might include getting feedback on project ideas or even looking for other members of the SANS community to form a team.

Acknowledgements

This competition is organized by James Thompson, Mark Thornton, and Luke Chang and is supported by The Dartmouth Consortium for Interacting Minds and an NSF CAREER Award 1848370 to Luke Chang. cim_logo

next

Accessing Data