- Newly developed AI predicts blood test for COVID-19 faster than standard tests
- Research could reduce need for future lockdowns
- Research team wins 1st prize in UK competition
A University of Sheffield research student has helped develop Artificial Intelligence (AI) to predict COVID-19 from standard blood tests two weeks earlier than current, existing tests.
This new AI could increase testing capacity and reduce the need for large scale lockdowns by spotting new waves and outbreaks before they develop, allowing for localised lockdown and social isolation.
The AI solution has the ability to input patient data for hundreds of millions of people and predicts the likelihood of outcomes such as how the disease will progress and the success of specific treatments. This could prevent new waves of COVID-19, including mutations. Mutations may not be picked up by current antibody tests, this could bring the reproduction number (the rating of a disease's ability to spread) to less than one. If the reproduction number is less than one this means a disease will eventually stop, as not enough new people are being infected to sustain the outbreak.
The AI could also detect and help isolate centres of future pandemics, as well as finding other information that is hidden in medical big data. In addition, by enabling lockdowns to be lifted the team’s solution will reduce the negative economic impacts of the current crisis and so allow faster recovery after the crisis.
Researcher Michail Mamalakis is currently studying his PhD in the Department of Computer Science at Sheffield and is part of a wider research team from the universities of Brighton, Bristol, Glasgow, Lincoln and Oxford. Michail’s contribution to the research was literature research, training and testing of artificial neural networks for detection of COVID-19 cases based on biomarkers (a measurable indicator of the severity or presence of disease), and statistical analysis of blood measurements of patients per fourteen days. Michail is also working on applying medical imaging methodologies as a potential detection tool of COVID-19 cases. Specifically, he works on COVID-19 cases detection, by using deep learning networks based on x-rays and CT images of lung regions.
The research team won first prize in the #AIvs.Covid-19 hackathon (which is a problem solving event for computer programmers) and took their proposal to the #EUvsVirus European-led Hackathon which took place in April.
Michail said:
“We had less than 4 days to analyse the data and train our models, but we succeeded and delivered a workable artificial intelligence model with high accuracy to predict if an individual is positive or negative of covid-19.
Our next step is to apply our ANN (artificial neural network) model in datasets across the world with a higher number of patients. Moreover we would like to study different ways to predict the severity level of the disease such as medical image analysis. We hope that with our work it will be possible to create patterns to avoid a new global pandemic.”
Editors notes:
Michail Mamalakis PhD supervisor is Professor Richard Clayton, Director of Training at Insigneo. Michail Mamalakis works mainly with deep learning, ML, image analysis, computer vision, C/C++ and Python.
The Research team is led by Dr Mark Baker, a Tech Start-up CEO and includes Dr Louise Mackenzie (University of Brighton), Dr Surajit Ray (University of Glasgow), Shelley Taylor (Programme Manager), Dr Joanne Kitson (University of Bristol), Dr Abhirup Banerjee (Oxford University), Dr Bart Vorselaars (University of Lincoln).
Details of the proposal:
https://medium.com/@pauldowling/accelerating-scientific-collaboration-in-real-time-e1f682f54c87
https://devpost.com/software/europa-mapp-predicting-stopping-covid19-waves-pandemics