We are told time and time again that the key to battling the coronavirus pandemic is testing. Yet even in developed countries, governments are struggling to provide enough tests to keep up with the daily number of cases.
The standard COVID-19 test uses PCR (polymerase chain reaction) to amplify and detect viral RNA, but can be time consuming and expensive to implement on a large scale. More recently, antibody tests have become available on the market, however the accuracy of some of these kits leaves much to be desired.
Machine learning could allow x-ray images to be used as an additional way to identify COVID-19 cases. Where a human would struggle to accurately differentiate between the chest x-ray of a COVID-19 patient and those suffering from another lung condition, artificial intelligence can ‘learn’ to identify COVID’s distinct features. In a study posted last week in MedRxiv, a machine learning algorithm was able to make this distinction with around 80% accuracy, providing a proof-of-principal for the use of AI in COVID-19 diagnosis. This approach could also help predict outcomes and optimal treatments for patients, by comparing their chest x-rays to those of previous cases.
It is not yet clear exactly how this approach would fit into the current testing strategy. X-rays would not be appropriate to diagnose mild or asymptomatic cases, and their use is limited to facilities with the necessary equipment, making them unsuitable for mass testing. Rather, they could be used to free up conventional tests for mass testing, or to provide an additional level of accuracy to current testing procedures.
In order to make sufficiently accurate predictions, a machine learning algorithm would need to be ‘trained’ on thousands of lung images, but acquiring this quantity of data is challenging. For example, the AI firm Zegami currently has just over 200 such images, but needs around 10 000 for their model, and is appealing to governments for assistance.
The fight against COVID-19 is a global one so we have written to the health ministers of a number of countries asking if they can help us with the development of our new platform. As soon as we have enough x-rays it will be fully up and running and we hope ready to play a key role in supporting medical and technical professionals in their battle with this disease.Roger Noble, CEO and founder of Zegami
A deep learning algorithm using CT images to screen for Corona Virus Disease (COVID-19): https://doi.org/10.1101/2020.02.14.20023028
AI firm asks for COVID-19 x-rays to develop new machine learning model: https://www.med-technews.com/news/ai-firm-asks-for-covid-19-x-rays-to-develop-new-machine-lear/