How AI, Big Data and Machine Learning can also be used against the Corona virus

covid_oakridge, Screenshot from “Researchers enlist Summit supercomputer to combat coronavirus” video by Oak Ridge National Laboratory

In the first part of this blog post we focused on mapping the coronavirus pandemic and we looked at how Big Data can be used for forecasting. In the second part, we are now looking at specific individual applications in areas such as the treatment of the virus, the diagnosis of sick people, but also at robots or drones that could potentially perform useful work in which people would be exposed to a risk of infection. Some of these approaches are certainly worthy of critical examination and are not without their problems, for example with regard to the protection of personal data. We would like to encourage you to form your own opinion, but also provide a basis for discussion. But see for yourself:

Finding Cures and Vaccinations

One of the first measures taken after the problem of the SARS CoV-2 pathogen became known was that researchers around the world began to use artificial intelligence and supercomputers to search databases for potentially effective, already known substances that could be used against the virus.

Two organizations that we would like to mention here by way of example are, on the one hand, the Oak Ridge National Laboratory in the USAMachine Learning Institute at JKU Linz

The first one has summarized very clearly in a video what they are working on exactly and why:

Dave Turek, IBM Cognitive Systems Vice President of Technical Computing, explains that using the IBM supercomputer shortens the first phase of testing potentially effective substances from several years to just a few days. 800 different chemical compounds were reduced to 77 promising compounds in just a few days. However, the second phase, the clinical tests, take just as long – at least 12 to 18 months are expected.

“Before, it was like throwing darts in the dark, blindfolded, on a dark disc. Now we have the arrows in our hands, we can see the target and try to hit it,” says Turek.

The same was done in Linz, among other places: Linz AI researchers are also testing medical substances for SARS-CoV-2 efficacy. 30,000 promising molecules were listed in a freely accessible database, these were again prioritized within the list. With the help of this preliminary work, a drug against the corona virus should be found as quickly as possible.

“Based on our experience with the use of artificial intelligence in the development of active pharmaceutical ingredients, we have programmed an artificial intelligence with the search for promising drug approaches against the new corona virus. In the fight against the corona virus, we make the data thus obtained available free to the international community. It is our urgent concern to support those who are working on medicines and active ingredients with our knowledge. The clearer we know where to look, the faster research can be done.”

Günter Klambauer, Institute for Machine Learning at Johannes Kepler University Linz
Promising molecules, photo: Institute for Machine Learning / JKU Linz

One project that we would like to mention, even if it is not directly related to AI, is the “Pressure Cuff Emergency Ventilator” of the ACZ Labs. Aisen Caro Chacin, artist, engineer and already represented several times at Ars Electronica, together with Christopher Zahner, has succeeded in “hacking” the components of a ventilator. In other words, using materials that are part of the basic equipment of medical supply cabinets, they have succeeded in recreating a pressure cuff emergency ventilator in order to increase the capacity of ventilators in case of emergency.

“The design consists of a blood pressure cuff that can be activated by medical air and vacuum connections in any patient room. Air inflates the cuff, compressing an ambu-bag, and the vacuum allows the air to escape quickly; this is repeated to achieve the desired 300-600 tidal volume within 20-30 rpm. This mechanism is controlled by an Arduino UNO, an H-bridge and two electromechanical vents. This design allows medical staff to improvise treatment if more ventilators are needed”.

In order to approve the device for medical use, an exemption certificate is requested. The cost of the non-medical devices is $105.25, which would allow for a doubling of ventilation capacity.

Aisen Caro Chacin and Christopher Zahner, photo: ACZ Labs

Diagnosing infected persons

In addition to RT-PCR tests, some doctors and scientists use computed tomography to detect and diagnose covid 19 infection or to detect pneumonia due to covid infection. Dutch-based radiologist Erik R. Ranschaert, who calls himself an “AI Believer”, has therefore launched a research project together with colleague Laurens Topff to implement AI in CT scans.

„We are going to train an algorithm with CT images with the intention of making it easier to diagnose COVID-19 with CT and to enable a correct assessment of the degree of lung damage. This will be an exciting European project and we hope to have results within a month. We are ready to start. Fingers crossed!“

Erik R. Ranschaert in a mailout

One possible application of AI in terms of diagnostics concerns temperature measurements on people in order to detect disease with Covid-19 due to elevated temperature. The use of machines in this area saves – so the basic idea of using AI – human labour where it is not absolutely necessary and also represents a risk of infection. In addition, machines can carry out trained processes faster than would be possible for humans.

The technology is not only used for access control, in hospitals or at airports; China has also equipped the police with helmets that make it possible to use temperature measurement to detect infected persons in the city. Debates on fundamental rights cannot be ignored here.

Simplifying Research

Anyone who has already done research knows how time-consuming it is to research and prepare the state of current research on a particular topic and to put the results in relation to each other. The time pressure is even greater when it comes to research on a new type of virus, where every day counts. AI can help here as well.

The Allen Institue for AI, together with leading teams of researchers, has published the COVID-19 Open Research Dataset (CORD-19). This is a free resource containing over 29,000 scientific articles, including over 13,000 full-text articles, on COVID-19 and corona viruses in general. Using Natural Language Processing, researchers can quickly update and process the latest research results. The corpus is updated weekly as new research results are published in peer-reviewed publications and archive services such as bioRxiv, medRxiv and others.

The London-based company BenevolentAI does a similar job: “We have developed Benevolent Platform™ – a leading platform for computer-aided and experimental research that enables our scientists to find new ways to treat diseases and personalize medicines for patients”.

AI against Fake News

A problem that has also been enhanced by the Corona crisis is that of fake news, misinformation on the net and rapid spread through social networks. In March, companies such as Google, Facebook, YouTube and Co. announced in a joint statement that they would try to combat misinformation on the net more quickly using AI. At the same time, they also warned of a higher susceptibility to errors: all employees are in the home office and AI is simply not as accurate as the people who work with it and control it.

Robots and Drones in Action

What is already reality in China and other Asian countries was not imaginable for many European countries before the corona crisis and is now partly reality: The increased use of drones to comply with the exit restrictions, for example.

In China and the USA, robots are also already being used in cleaning, nursing and caring for sick and therefore infectious people. Drug transport by means of drones is still a dream of the future.

Finally, Something Completely Different:

AI is also used, as it is known by now, to imitate the voices of famous people, for example, as it was done in this short clip.

You have discovered more interesting research projects around AI and Corona or are working on one yourself? We’d be happy if you let us and the Ars Electronica community know about it!

, ,