Call for Demonstrations

Increasing numbers of Machine Learning-based Software as Medical Devices are approved by organizations such as US FDA, China NMPA, or EU CE among others. As the ML4H field continues to mature and differentiate, there is a growing need for an interface where assumptions prevalent in ML4H research can be validated against the challenges, solutions, and maturity of real-world ML4H tools. The ML4H Demo track aims at submissions that demonstrate real-world applications of ML4H technologies, bridging the gap from proof-of-concept to practical utility.

Accepted submissions will be non-archival and have the opportunity to present their live demo on the day of the event alongside the main poster sessions. Select outstanding demos will be given spotlight talks. Reviewing for the Demo Track will be single blind.

Important Dates

Aug 1st: Submission site opens
~~Oct 6th AoE~~ Oct 31st AoE: Demo submission deadline
Nov 20th: Decisions released
Dec 15-16th: In-person event

Submission Instructions

Submission Site: https://openreview.net/group?id=ML4H/2024/Demo_Track

Each Demo submission must contain the following two components:

1. Spec Sheet

A short writeup (max 2 pages, excluding references) describing the ML4H tool, technology, and application, typeset using the following LaTeX template. The Spec Sheet should contain the following information:

Introduction: What is the problem that the tool is trying to solve? Why is it important? How would an ML-based solution solve the problem? This should cover aspects of the tool including intended use, intended patient population, principles of operation and conditions of use.
Method: How does the tool work technically? What is the ML technology behind it, and how was it developed? Here, you should describe the ML algorithm, the model architecture, the training data and procedure, and the deployment pipeline.
Results: Explain the state of the tool, and any measurable outcomes that it has had during its deployment. How has the tool performed during deployment, versus during development? How many users or patients does the tool impact, and how widely used is it?
Discussion: What were some challenges of developing, deploying, or operating the tool? What were some lessons learned? How could the tool and its deployment have been improved in hindsight, and how do you plan on improving it in future?

As reviewing is single blind, the spec sheet should not be anonymized. The spec sheet will only be viewed by the ML4H Demo Review Committee and will not be made public.

2. Demo Video

A link to a video (at most 2 minutes long) demonstrating the tool in use with a voice-over description. You may assume that the viewer has read the spec sheet prior to watching the video. The video will only be viewed by the ML4H Demo Review Committee and will not be made public. Any PHI or confidential information should be blurred or omitted.

As OpenReview is unable to host large video files, you should first upload the video to a platform such as Dropbox, Google Drive, OneDrive, Vimeo, or YouTube (unlisted is okay), and provide a link to the video in the submission form. Any common video format is acceptable (e.g. MP4, MOV, WMV, AVI). Any submission which does not have a working link to a demo video will be desk-rejected. The demo video will only be used to assess the tool itself, and any fancy editing and VFX (or lack thereof) will not impact the assessment.

Please let us know, however, if there are any extenuating circumstances which prevents you from making and submitting a video.

Selection Criteria

All submitted demos will be evaluated based on the following selection criteria:

Relevance to the ML4H field
Maturity of the tool or project
Significance and impact of the tool
Quality and clarity of the submission
Discussion of challenges and lessons learned

All submissions will undergo a review process by the ML4H Demo Review Committee to uphold the selection criteria and assess the maturity and fit of the submitted demos.

Registration Information

To promote community interaction, at least one presenting author of accepted works must register for the event. Registration details are forthcoming.

Contact Us

Please direct questions to info@ml4h.cc and follow us on Twitter at @symposiumml4h.