AI Health Data Challenge

Home Teaching Research Publication About Me

AI Health Data Challenge

Organized by:
School of Information at UT Austin
Dell Medical School at UT Austin
Critical Data at MIT



Sponsored by:
Suit Endowment Fund
Mary R. Boyvey Dean's Excellence Fund


1st prize: $1,500; 2nd prize: $500


Get our flyer


Are you passionate about AI and healthcare? If yes, read on for the details.

It is our goal to bridge disciplines and promote data-driven, evidence-based care. We are looking for participants from diverse backgrounds with skills in information science, computer science, and life sciences, to leverage their collective knowledge as a team and create an app or a data analytics tool/package for mining large datasets of electronic medical records. The app or tool should produce valuable knowledge to support clinical decision making and present it through an effortless and immediate interface.

What benefits ME/Why do I have to care?

Everyone is busy, so spending time wisely is critical. Joining this event can bring you the following benefits:

  • Building a strong resume: this challenge is based on real-world large scale datasets in healthcare. It will provide you hands-on experience to better understand issues in healthcare and how AI can help.
  • The power of data and design: The winning innovations in industry are based on the combination of data and design (e.g., Apple, Tesla, and more). Unfortunately data and design are divided into distant groups. This challenge provides a unique chance to merge these distant groups so that you can team up together to uncover the magic of data and design.
  • Solving burning problems in Healthcare: Healthcare is now the US's No.1 employer with 1 in 8 americans working in this sector. The U.S. Bureau of Labor Statistics states that there are 16 million medical-related jobs out there and the profits this sector generates reaches to $2.7 trillion a year. Solving burning issues (e.g., improving quality of care, delivering care intelligently, and reducing cost by using AI technologies) can have an immense impact in this country and worldwide.
  • Meeting peers: During this challenge, we encourage participants to work in teams to truly bring data and design together. So, you will meet like-minded friends and creative souls. We use slack to boost the communication and help you form teams.
  • Networking: On our judging day (April 3, 2020), you will meet our judges who are leaders and experts in academia and industry. You will also meet other teams participating our challenge. It will be a great networking event.
  • Knowing more details: All submissions of the finalists will be shared with our participants.

Data

This challenge is based on MIMIC datasets. Currently MIMIC has three datasets available. Since the datasets have to be treated with appropriate care and respect, researchers seeking to use these datasets must formally request access. Access is granted for an individual person. So even if you are working in a team, each team member needs to get the separate permission. Prior to requesting access to MIMIC, you will need to complete the CITI online course “Data or Specimens Only Research”. It usually takes 1-2 weeks to get permission. Please plan accordingly. Details about the datasets and access are the following:

  • MIMIC ICU dataset: MIMIC is an openly available dataset developed by the MIT Lab for Computational Physiology, comprising deidentified health data associated with ~60,000 intensive care unit admissions. It includes demographics, vital signs, laboratory tests, medications, and more. Details on how to request access are: https://mimic.physionet.org/gettingstarted/access/
  • MIMIC eICU dataset: The eICU Collaborative Research Database is a large multi-center critical care database made available by Philips Healthcare in partnership with the MIT Laboratory for Computational Physiology. Details on how to request access are: https://eicu-crd.mit.edu/gettingstarted/access/
  • MIMIC Chest X-ray Image and Report dataset: The MIMIC Chest X-ray (MIMIC-CXR) Database v1.0.0 is a large publicly available dataset of chest radiographs with structured labels. The dataset contains 371,920 images corresponding to 224,548 radiographic studies performed at the Beth Israel Deaconess Medical Center in Boston, MA. The dataset is de-identified to satisfy the US Health Insurance Portability and Accountability Act of 1996 (HIPAA) Safe Harbor requirements. Protected health information (PHI) has been removed. The dataset is intended to support a wide body of research in medicine including image understanding, natural language processing, and decision support. Details on how to request access are at:https://mimic.physionet.org/gettingstarted/access/.(Notes: once you get the permission for MIMIC ICU dataset, you can use the same permission to download MIMIC-CXR dataset).

Team

Need teammate(s)? You can join our Slack channel to find team members who have registered for the same. Visit https://invite-ut-ai-data-challenge.herokuapp.com/ to request an invite to the Slack channel. You can also use this to contact us regarding any questions about the challenge. In case you have formed a team amongst other participants already, only one member per team needs to fill out this form. You will have a chance to fill out the details of other team member(s) on the next section of this form. Please fill out your details on this Google Form.

Final Deliverables

Please submit the following items before 11PM Central Time March 2nd, 2020 to this UT Box folder (Note: submit only one zip file containing everything and filename as the submitter's full name):

  • app/tool: the link to your GitHub
  • A 2-5 page report on the details of how you built the app. This report should contain the details of your methods, and screenshots of your app/tool.
  • 5-minute video about your process

Deadlines

  • Register at the Google Form by December 1, 2019
  • Download dataset no later than November 29, 2019
  • Final submission due by 11pm CST on March 2, 2020
  • Announcement of finalists by March 22, 2020
  • Presenting, judging, and announcing final winners: April 3, 2020

How to win

Your submission will be shortlisted by UT faculty by March 22, 2020. The top teams will be invited to participate in a day-long seminar on April 3, 2020, where each team will present their work to a jury of healthcare professionals (e.g., faculty, students and industry experts). The judging criteria are:

  • Usability: Is your app/tool useful to address some healthcare issues? Is your app/tool easy to use?
  • Novelty: What are NEW features in your app/tool? Are there any creative and exciting things in your app/tool?
  • Reproducibility: Can others reproduce your app/tool?

Contact Us

If you have any questions, please feel free to reach us at ying.ding@ischool.utexas.edu