2007 PBAIC Overview

Prizes: 1st $10,000; 2nd 5,000; 3rd $2,000
Reviewers Choice: $5,000 (US dollars)
Draft Announcement, details subject to change

Modern advances in brain imaging provide the foundation for interpreting relationships between fMRI (functional Magnetic Resonance Imaging) brain activation signal and subjective experience. The purpose of this competition is to challenge multiple groups to use state-of-the-art techniques to infer subjective experience from a rigorously collected set of fMRI data associated with dynamic experiences in a virtual reality environment with a quantitative metric of success. Ideally, the analysis of brain activation patterns can characterize what an observer experienced while immersed in the VR environment (e.g., was there a face present on the screen? What was the subject instructed to do? Was there background music? Was the subject being followed by a vicious rottweiler? How anxious was the participant?). To advance the methodology and assess the state of the science, the University of Pittsburgh, will provide a competitive prize for the groups showing the most effective methodologies. A board of scientists has designed the test set of activation images.

Groups from all nations and disciplines are encouraged to participate. There are no restrictions on publication of the data other than crediting the data source and subject confidentiality. Awardees will be required to describe their methods and either present the methods or provide written descriptions at the Competition Workshop at the Organization for Human Brain Mapping, June 10-14, 2007 in Chicago, Illinois, USA.

The competition will involve analysis of fMRI data of individuals experiencing three 20-minute segments of immersion in a virtual reality environment with extensive behavioral time vector ratings of multiple categories representing what they perceived/experienced on screen (i.e. human faces, tools, arousal, etc.). Subjects were instructed to complete search tasks in the environment (looking for weapons, taking pictures of people with piercings, or picking up fruits in a pre-specified order). 3T EPI fMRI data from multiple subjects in 3 20-minute sessions will be provided. For the first two sessions, video of the subject’s path through the virtual reality environment, along with 20 minutes of continuous fMRI data and feature ratings will be provided. For the third clip the ratings will not be provided. The goal will be to predict feature rating vectors for each of the segments. Accuracy of predictions will be determined by correlating predicted behavioral ratings with the empirical ratings on a second by second basis. The three most successful groups will be asked to present brief summaries of their mapping techniques.

Prizes will be awarded ($10,000 1st; $5,000 2nd; $2,000 3rd; $5,000 Reviewers’ Choice) at the Organization for Human Brain Mapping meeting, June 10-14, 2007 in Chicago, Illinois, USA. The data sets will be publicly available as a test bed for future research teams. Details regarding the meeting can be obtained at: http://www.humanbrainmapping.org

Brief details of the competition:

Groups must register to receive access to the data. There are no costs for registering. Groups must register before May 1, 2007. Registered groups will receive emails on issues that might assist them.
Groups must abide by the rules to predict experience based only on the brain imaging data for the final clip though they can calibrate methods on an individual subject basis for the first two movies. They must also abide to rules of subject confidentiality (e.g., never showing images showing the face of the subject) and if they use the data in publications cite the EBC research effort. Groups must agree not to redistribute the Pittsburgh data set to other groups (though they can direct people to get the data from Pittsburgh authorized data sites). Groups agree to allow the competition group to run statistics on the submitted predictions and to report analyses of those results (e.g., reporting of the accuracy of the submitted entries). All decisions by the scientific judging panel will be final.
Each group is allowed unlimited opportunities to work with data sets A & B. However they are allowed only three opportunities to submit their predicted behavioral vectors scored on segment C. (Data will be made available for analysis after the competition is completed. )
Final submissions of data by May 21, 2007 - 12:00PM EDT.
Winning groups must disclose the methods used to make the predictions and be willing to either make a presentation at OHBM or submit a short paper/talk/poster describing the methods. The group must answer any questions regarding the methods used coming from the judging panel. They will also provide required data of their results to enable meta analysis of the submission efforts. The top contenders will be required to upload predictive data voxel weighting data (providing a monotonic metric of how each voxel contributed to any specific feature score. They will also be required to provide pseudo-code (or data flow diagrams) describing in detail the analysis steps employed to perform the rating prediction and Sensitivity analysis of their technique (e.g., how if they turn off spatial filtering how much it reduces the predictive accuracy).
No groups having access to the segment C data set behavioral ratings will be allowed to enter the competition.

Complete details of the competition and the procedures that will be used can be found on the competition materials page.

Competition Overview