[Paper Review] Hi Sigma, do I have the Coronavirus?: Call for a New Artificial Intelligence Approach to Support Health Care Professionals Dealing With The COVID-19 Pandemic

Brian Subirana, Ferran Hueto+11 more|arXiv (Cornell University)|Apr 10, 2020

COVID-19 diagnosis using AI63 citations

TL;DR

The paper proposes an open, collective AI approach using cough audio transfer learning to diagnose COVID-19 and outlines validation plans across multiple sites, plus tools for longitudinal testing and ICU prioritization.

ABSTRACT

Just like your phone can detect what song is playing in crowded spaces, we show that Artificial Intelligence transfer learning algorithms trained on cough phone recordings results in diagnostic tests for COVID-19. To gain adoption by the health care community, we plan to validate our results in a clinical trial and three other venues in Mexico, Spain and the USA . However, if we had data from other on-going clinical trials and volunteers, we may do much more. For example, for confirmed stay-at-home COVID-19 patients, a longitudinal audio test could be developed to determine contact-with-hospital recommendations, and for the most critical COVID-19 patients a success ratio forecast test, including patient clinical data, to prioritize ICU allocation. As a challenge to the engineering community and in the context of our clinical trial, the authors suggest distributing cough recordings daily, hoping other trials and crowdsourcing users will contribute more data. Previous approaches to complex AI tasks have either used a static dataset or were private efforts led by large corporations. All existing COVID-19 trials published also follow this paradigm. Instead, we suggest a novel open collective approach to large-scale real-time health care AI. We will be posting updates at https://opensigma.mit.edu. Our personal view is that our approach is the right one for large scale pandemics, and therefore is here to stay - will you join?

Motivation & Objective

• Motivate the development of a new AI paradigm to aid health care professionals during the COVID-19 pandemic.
• Propose transfer learning on cough audio as a diagnostic signal for COVID-19.
• Advocate for open, real-time data sharing and crowdsourced data collection to accelerate validation.
• Describe planned clinical trials and multi-site validation in Mexico, Spain, and the USA.

Proposed method

• Use AI transfer learning trained on cough recordings to create diagnostic tests for COVID-19.
• Plan validation through clinical trials and three international venues.
• Propose longitudinal audio testing for confirmed stay-at-home patients to inform hospital contact guidance.
• Suggest a patient-centric forecast tool for ICU prioritization incorporating clinical data.
• Advocate for an open collective approach to large-scale real-time health care AI.

Experimental results

Research questions

• RQ1 Can cough audio obtained from mobile devices be used to accurately diagnose COVID-19 using transfer learning?
• RQ2 What are the feasibility and benefits of validating AI-powered cough diagnosis across multiple international sites?
• RQ3 How can longitudinal audio data inform treatment decisions and hospital resource allocation for COVID-19 patients?
• RQ4 What is the potential impact of an open, crowdsourced data model on rapid AI development during pandemics?

Key findings

• AI transfer learning on cough recordings can yield diagnostic tests for COVID-19 according to the authors’ claim.
• A clinical trial and international deployment plan is proposed to validate the approach.
• A longitudinal audio testing concept is described for stay-at-home patients to guide hospital contact decisions.
• A success ratio forecast test for critical patients is proposed to prioritize ICU allocation when combined with clinical data.
• The authors advocate an open collective approach to large-scale real-time health care AI and data sharing.