logo background
Multisite resting-state fMRI Initiative (PsyMRI)
  • Slide 1
  • Predictive Analytics in Mental Health Competition (PAC)


    Thank you for your interest in the Predictive Analytics in Mental Health Competition (PAC).

    This year, we invite you to develop tools to classify patients suffering from Depression and healthy individuals based on resting-state functional Magnetic Resonance Imaging (rs-fMRI) data. The goal is to provide models which correctly label previously unknown datasets as stemming from either depressive or healthy individuals. Information on how to obtain the resting-state fMRI data you can use to train the model (henceforth called "the training data") will be provided on the next page.

    All models submitted before the official end of this year's PAC will be tested on fresh data of depressive patients and healthy participants (henceforth called "the test data") which is currently accessible to the organizers only. The team submitting the model classifying most of these new, previously unseen datasets correctly will win this year's PAC. In case of a draw, all teams achieving the highest number of correctly classified datasets will win the PAC.

    The data for this competition was provided by the PsyMRI Consortium – a organization aimed at a multicenter-based utilization of resting-state magnetic resonance data from member sites with a focus on major depressive disorder (MDD) and neurodegenerative diseases. Consortium members from 11 institutions around the world provided de-identified, HIPAA compliant rs-fMRI datasets from 436 subjects with and without depression.

    By participating in this year's PAC, you agree

    1. to use the training data provided by the PsyMRI Consortium for no purpose other than the model construction relevant for this year's PAC.
    2. not to publish any model trained fully or in part on the training data without written permission of the PAC organizers.
    3. not to distribute the training data to people who are not members of your PAC team.
    4. to submit only one model per PAC team.
    5. that only models submitted before September 11th, 2016 will be considered.
    6. that the organizers will use your model to classify the test data. To enable this, you furthermore agree to provide additional software needed to derive predictions from your model if necessary.

    Note that the organizers will not use your model for any purpose other than to classify the test data unless written permission has been obtained from you.

    If you have any questions, do not hesitate to contact the PAC organizers (Tim Hahn, Hahn@psych.uni-frankfurt.de).

    Good luck and have fun!




    In cooperation with:

    Conference on Resting State and Brain Connectivity 2016


    BrainHack Vienna 2016
    Copyright © 2016 Canlab | Impressum | Design: templatemo - Conquer