Fujitsu Cough in a Box
Released: April 2021
A team of researchers from The Alan Turing Institute and Royal Statistical Society, commissioned by the UK Health Security Agency have discovered that cough-analysing algorithms do a poor job of diagnosing COVID-19.
The team found that even the most accurate cough-detecting model performed worse than a model based on user-reported systems and demographic data, such as age and gender.
One such model was Fujitsu’s Cough in a Box, an app funded to the tune of GBP 100,000 in 2021 by the UK’s Department of Health and Social Care to collect and analyse audio recordings of COVID-19 symptoms.
In October 2020, a team at the Massachusetts Institute of Technology (MIT) had found that, when new cough sounds were introduced, the algorithm accurately identified 98.5% of coughs from people who were confirmed to have COVID-19, including 100% of coughs from people with no symptoms.
The researchers had used technology that aimed to identify a signature noise, known as a biomarker, in a person’s cough that could be linked to a positive COVID-19 PCR test result. Their findings led to efforts to build an algorithmic-powered app to provide people with a cheap and easy method to test for COVID-19.
Operator: Department of Health and Social Care (DHSC)
Developer: Fujitsu; Formwize; Cloudsoft
Purpose: Diagnose COVID-19
Technology: Machine learning
Audio-based AI classifiers show no evidence of improved COVID-19 screening over simple symptoms checkers
Statistical Design and Analysis for Robust Machine Learning: A Case Study from COVID-19
A large-scale and PCR-referenced vocal audio dataset for COVID-19
Laguarta J., Hueto J.F., Subiran B. (2020) COVID-19 Artificial Intelligence Diagnosis Using Only Cough Recordings
News, commentary, analysis
Published: February 2023