IEEE Brain Data Bank Competition - Boston, MA
This is the last in a series of challenges and competitons sponsored by IEEE Brain Initiative in 2017 that explore various brain/neuro datasets. Results and final presentations are expected to be made at the Boston (Cambridge) event, December 9, 2017. NOTE: EVENT IS STILL ON AS SCHEDULED DESPITE WEATHER.
COMPETITION DETAILS: https://brain.ieee.org/news/call-participation-ieee-brain-data-bank-chal...
REGISTRATION FORM: http://bit.ly/2xYX40o
EVENT LOCATION: Charles River Analytics, 625 Mt. Auburn St., Cambridge, MA, USA
An audio bridge will be made available for you to listen though we cannot guarantee the quality of the audio. AGENDA [Updated on 12/5/17]: Download under 'DOCUMENTATION' (to the right)
AWARDS: 1st place - $800.00, 2nd place - $500.00, 3rd place - $200.00 (judging criteria based on data analysis, data visualization, and presentation and interpretation of findings)
SLACK CHANNEL: Please join this channel for answers to FAQs and to post additional comments or questions.
1) DOWNLOAD THE DATASETS NOW. Data is available. Complete access request form below.
2) DOWLOAD THE RECORDED WEBINAR if you missed this which took place on November 21. Slides and recording available for download under 'DOCUMENTATION' (to the right).
3) Project title is due December 1st to secure a presentation slot on December 9th.
4) Final presentation on December 9th, from 9:30 am to 4 pm. We recommend no more than 5 slides. Teams have 5 minutes to present and 5 minutes of questioning from the judges. All must bring IDs. If you are not a U.S. citizen, please bring your passport. COMPETITION DATA DESCRIPTION:
Datasets are provided by the University of Illinois at Urbana-Champaign (UIUC) via funding provided by the Intelligence Advanced Research Projects Activity (IARPA) under the Strengthening Human Adaptive Reasoning and Problem-solving (SHARP) program. Please complete the form below to request access to the datasets.
The provided data comes from a broader intervention-based longitudinal study, designed to test the efficacy of interventions designed to enhance fluid intelligence. This specific intervention involved 48 training sessions of an adaptive visuo-spatial and change detection task. We provided both behavioral measures and neuroimaging data at both pre-intervention and post-intervention. We also provided summary and session-level descriptions of the training data collected during the intervention.
We provided pre- and post-test measures of two standardized tests of fluid intelligence: Figure Series, and the analogical reasoning portion of the Law School Admissions Test. For both tests, we provided item level accuracy (coded as 1=correct, 0=incorrect) and reaction time (seconds). Behavioral measures at each timepoint are provided in a *.csv file, where the 25 rows represent participants and the 25 columns (for LSAT) or 30 columns (for Figure Series) represent test items. We also included three measures of processing speed: Letter Comparison, Pattern Comparison, and the Digit Symbol Substitution Task. We also included basic demographics, such as age, sex and years of education.
We provided intervention training data from three versions each of a Visuospatial and Change Detection task. These data are provided in a *.csv file, containing the average difficulty level across 48 training sessions for each version of the tasks.
We provided three types of neuroimaging data for each participant. First, we provided the structural T1 scan, a high-resolution image of the brain. Each *.nii.gz file is about 13MB in size, with a 0.9mm resolution. Second, we provided functional neuroimaging data acquired at rest. Resting state data are provided in the form of a processed functional connectome, a *.csv format file containing pairwise functional connectivity values from 256 cortical regions. We also provided the pre-processed and filtered resting state scans in *.nii.gz files, about 350MB in size each. Finally, we provided probabilistic tensor and tractography data from a Diffusion Tensor Imaging (DTI) scan. For each subject, we provided *.csv files describing the probabilistic fiber paths of white-matter structural connectomes between 68 cortical regions. We also provided *.nii.gz files, about 1MB in size each, containing probabilistic tensor components describing anisotropic diffusion in the brain. All files are provided at pre- and post-intervention.
The full dataset is 11.29GB. The dataset can be downloaded in full, or in more manageable chunks separated as:
- Behavioral and analyzed fMRI/DTI data
- Structural T1 images
- Raw resting-state fMRI data (divided in 5 file groups)
- Raw DTI data
NEW STANDALONE FILE UPLOADED ON NOV 8, 2017 (TestingDataAllSubs.csv) - This spreadsheet summarizes the educational attainment and behavior data for each subject. The education level is identified as follows: 'No high school', '1', 'Some high school', '2', 'High school graduate', '3','Some college', '4', 'College graduate', '5', 'Some post-graduate', '6', 'Master''s degree or higher', '7' .
Participants must use the data, and any solution derived from the data solely for the purpose and duration of the Competition, including but not limited to privately sharing data outside of teams or for publication, unless provided express permission by UIUC and the IEEE Brain Initiative. If permission is granted to publish or release data in any capacity, acknowledgment must be given to UIUC and the IEEE Brain Initiative in terms provided by them.
The goal of this challenge is to encourage thoughtful investigation and discussion around enhancement of human intelligence and strategies for rigorous investigation of these topics.
Participants can investigate any question(s) about the data that they wish. The following guiding questions are provided as a jumping off point:
- Can intervention training predict changes in pre- and post-intervention performance?
- How do measures of brain connectivity and structure relate to performance in intervention training?
- How do measures of brain connectivity and structure relate to pre- and post-intervention performance?
- How well can we predict post-intervention performance? Beyond pre-intervention performance, what factors serve as the best predictors?
- Do performance changes on the two training intervention tasks differ? How do they each relate to changes in pre- and post-intervention performance and/or imaging data?
Awards will be presented. To be considered for an award, final presentation must be given December 9th in Boston (Cambridge).
We will accept recorded presentation if you are not able to attend. Please send your recording to firstname.lastname@example.org. However, please note that we are not set up for remote participation. We will play your presentation to the judges, but they will not have an opportunity to ask you questions.
If you have questions, please contact email@example.com.