Mike Schaekermann*, Edith Law*, Rebecca Fiebrink**
* University of Waterloo (Canada)
**Goldsmiths, University of London (UK)
The ICAD17 sonification competition is unique in many ways, and offers exciting opportunities to contribute to applied research. Its theme is "Turning Brainwaves into Sounds." The objective of the competition is to sonify sleep EEG (human brainwave signals captured during sleep), with the goal of creating mappings that help to distinguish signals associated with different sleep stages.
Important Dates
● Sign-Up Deadline: May 15, 2017
● Start of competition: June 5, 2017
● Submission deadline: June 18, 2017
● Announcement of winner: The ICAD 2017 Banquet
Background
The analysis of sleep EEG is essential for understanding sleep patterns and pathologies of patients. In current practice, medical technicians manually classify thousands of epochs of EEG data into 5 stages of sleep (Awake, N1, N2, N3, REM), a process that is extremely time-consuming and tedious. At University of Waterloo, we are developing new human-in-the-loop intelligent systems to more efficiently classify medical time series data. In a project called CrowdEEG (http://crowdeeg.ca), we study how to leverage crowdsourcing to tackle this problem.
Not surprisingly, it can be challenging to teach non-experts to do the expert-level analysis task typically handled by medical technicians. Our question, and the motivation behind this competition, is:
“By sonifying sleep EEG data, can we enable non-experts to classify the signal into sleep stages, or identify transitional points between sleep stages, simply by listening?”
Sonifying EEG is a non-trivial process, requiring expertise in sound engineering --- which is where you and the rest of the ICAD community come into play!
Objective
The goal is to maximize the within-class similarities (e.g., make all the REM sleep epochs sound similar) and between-class distances (e.g., make the REM sleep and Awake epochs sound different), so that the crowd of non-experts can easily distinguish between the 5 classes, and identify transitional points between sleep stages.
Procedure
Participants will be provided with (1) a web-based sonification tool that allows them to construct a mapping from time series to sound, (2) a dataset containing sleep EEG data, consisting of several types of biosignal time series, including brainwave activity.
Evaluation
Instead of relying on expert jury members, we will use a crowdsourced approach to determine the best mapping. Each submitted sonification will be tested on a large number of novices via the crowdsourcing platform Mechanical Turk, and the mapping with the highest average classification performance among novices will win the competition.
Prize
The winning entry will be announced at the end of the conference, and included in the final ICAD17 concert program. As a prize, we will award the winner a gift card worth USD $200 for an online store offering different products for sound designers and sound engineers (e.g., http://www.rubadub.co.uk/).
How to Sign Up?
All conference attendees are eligible to participate. You will be able to indicate your interest in participating either as part of the registration process, or by email to the conference chair (icad2017chair@icad.org). Participants need to indicate their interest by May 15, 2017. Conference registrants who indicate their interest will receive email instructions to access the data and web interface.
Since this competition also serves the purpose of a research study, we will ask you to sign a consent form prior to participation. By signing the consent form, you will agree to participate as a research subject and grant us the right to collect and publish any data related to your interaction and experience with the sonification tool we will provide. The study will have official approval from the ethics boards at University of Waterloo and Goldsmiths and University of London.
While a number of participants in the 2017 ICAD conference expressed keen interest in the competition, they were unable to submit entries by the deadline. The organizers will be in touch with them about another deadline in the near future.
Organizers
The competition is organized by a group of human-computer interaction and machine learning researchers from the University of Waterloo (Canada) and Goldsmiths, University of London (UK), who are passionate about tackling health-related challenges by developing novel crowdsourcing solutions:
● Mike Schaekermann (mschaeke@uwaterloo.ca), University of Waterloo
● Edith Law (edith.law@uwaterloo.ca), University of Waterloo
● Rebecca Fiebrink (r.fiebrink@gold.ac.uk), Goldsmiths, University of London
Please reach out to the organizers or email icad17.sonification@gmail.com if you have any questions. We are very excited about this competition and look forward to your participation.