How AI can lead the way in battle against Covid-19
Artificial Intelligence can be used to predict which Coronavirus infected patient may develop severe respiratory disease
The rapid emergence and spread of Covid-19 is putting health systems under unprecedented pressure, with a surge in demand for treatment, diagnosis and contact tracing to curtail the spread of viral infections.
Although medical professionals play the central and most crucial role in this battle, scientists from other disciplines are also trying to develop, design and repurpose methods and technologies to help manage the crisis.
Computer scientists are among those who can play a significant role in the battle against Covid-19.
Artificial Intelligence (AI) is capable of performing human-like tasks and could be used in several different areas in the campaign against Covid-19.
AI can be used to predict which Covid-19 infected patient may develop severe respiratory disease. Decision-support tools equipped with AI capabilities may be used to predict which patients are likely to develop severe symptoms. NYU Grossman School of Medicine and Courant Institute of Mathematical Sciences at New York University developed a tool that can detect three features, liver enzyme, myalgia and haemoglobin, to be the most accurate predictive of severe disease.
AI is also providing a preliminary and temporary buffer for non-critical patients to be consulted, diagnosed and treated so hospitals and clinics can efficiently work on more urgent cases. Examples of such systems are DoctorLink or Babylon Health that use an AI system to perform a preliminary diagnosis. These tasks are usually performed by making use of chatbots and advisory robots that are types of AI applications.
AI can also be used to support contact tracing, which is one of the key ways to limit an outbreak. Countries are using technologies such as GPS and bluetooth on mobile phones to collect data and then using AI-equipped analytics tools for contact tracing.
Radiography image analysis is a supplementary method to detect Covid-19. AI and, more specifically, machine learning experts have been working on designing and developing tools to automatically detect pneumonia from both X-ray and CT images for more than a decade. The success of those tools led computer scientists to apply a similar technology to X-ray and CT (computed tomography) images to perform a preliminary detection of Covid-19.
Right now, multiple projects and research studies are racing to investigate the efficacy of these types of methods. Some have used previously designed models with new datasets (confirmed infected Covid-19 images) while others have developed a completely new network specifically for this matter.
One area being examined by researchers at Lero, the Science Foundation Ireland Research Centre for Software, and Cork Institute of Technology, uses AI to recognise Covid-19 and pneumonia in X-ray images and differentiates these from a healthy patient’s scan.
As a prototype, the team has developed a smart system that can diagnose pneumonia from healthy cases with greater than 95 per cent accuracy. This prototype is based on the latest machine learning technologies and methods.
An essential concern for tools like this one is ease of use, especially when it is going to be utilised by non-computer scientists - medical scientists or medical doctors. To address this, the prototype uses a simple online portal that is hardware and software independent and can be used with a reliable and secure internet connection (Preliminary implementation was done as part of an MSc project by Chand Sheikh, MSc in AI student at CIT).
The research is ongoing and focuses on two main areas, one of which is differentiating the Pneumonia caused by Covid-19 and Pneumonia caused by non-Covid-19 causes from X-ray images.
As CT images are more detailed and provide a three-dimensional aspect of the chest and lung, work is also being carried out to increase the capability to diagnose Covid-19 from CT images. This means patients who have Covid-19-like symptoms may be diagnosed with a higher confidence than where swab test is the primary testing approach.
Globalisation means people travel more, and this large scale movement between countries and continents can raise the risk of spreading diseases such as Covid-19. As we are witnessing, Covid-19 emerged in China and in just two months it had become a global pandemic.
Therefore, where international interactions are inevitable, the need to minimise and control the spread of such diseases is essential.
AI and computer-based techniques are playing vital roles to this end, and although some goals have been achieved, more academic research is essential.
Dr Farshad Ghassemi Toosi is a member of Lero, the Science Foundation Ireland Research Centre for Software. He is a lecturer in CIT’s Computer Science department.
Dr Ruairi O’Reilly (CIT) is a researcher in the area of Applied AI-based systems. In collaboration with University Hospital Limerick, University of Limerick and Lero, they are developing the Image-Based Pervasive Monitor and Management (IBPMM) Platform for Covid-19.