5 Ways that AI is Used to Combat Corona Virus [COVID-19]

Hey guys and welcome back. So today we are going to be talking about the world most trending topic… the Corona Virus and how Artificial Intelligence and Data science can be used to combat it.

This article will discuss the 5 main areas of AI research which are:

1. Detection or Diagnosis of the disease.
2. Tracking and Analytics of the spread of COVID 19 for Preventative Measures.
3. Reduction – Disease Control – How do we physically stop the spread.
4. Potential Cures, Treatments and Vaccines.
5. And just as important as the disease itself is halting the spread of Misinformation.

I’m not going to talk about what the disease is, how it works and current statistics because, there are a ton of news outlets, videos and sources that cover this in detail. I will just be focusing on How AI, Deep Learning and Data science is being used to combat COVID 19.

All the sources and references that I mention in this lecture will be in the links section in the video description. I’ve attempted to the best of my ability that these are credible sources such as news publications, journals and conference papers, and well-established blogs and websites.

I disclaim that this information is merely to inform you and that any part of the video should NOT be taken as medical advice. This is important, If you think that you may be infected, please… please go see a medical professional.

With that out of the way, I will also have the time stamp should you wish to skip ahead to a particular topic. Okay Let’s not waste any time and get right into it!

Detection & Diagnosis using AI

Before you can act on a particular problem, you first need to know when the problem has occurred. In this case you will need to detect or diagnose if a person has COVID 19 in both its active and passive state. The passive or incubation stage can span anywhere between 2-14 days and may easily go undetected.

Infrared Cameras are used to detect fever like symptoms, but it wont tell you whether it just a normal flu or corona virus. Jun Chen et al. uses a deep learning model for detecting corona pneumonia on high resolution computed tomography also known as a CT scan. There’s also a method of using saliva swabs to determine the infection type on the spot as well as corona testing drive thru’s are becoming more prominent around the world. Let’s take look at both these methods of diagnosis.

Swab Tests

So, swab test involves collecting samples from patients, by swabbing for their saliva, and testing them for specific kinds of molecules that appear in the genetic material of the new coronavirus. They work using a process called Nucleic acid detection. A nucleic acid Test (NAT) [1] according to Wikipedia is a technique used to detect a particular nucleic acid sequence and thus usually to detect and identify a particular species or subspecies of organism, often a virus or bacteria that acts as a pathogen in blood, tissue, urine, etc. In this case we use it to detect for Covid-19 [this long text could be a small text under the title text or screencapture of wikipedia page – See if you can find a video of NAT]. While these kits are said to be cheap, abundant and generally accurate, with the growing number of corona cases, test kits could be in shortage.

According to the Staits Times [2] there is a growing concern that NATs are producing a large number of false negatives, meaning that not all of those infected by corona return positive NAT’s which is pretty scary. An anonymous doctor at a major Wuhan hospital said that some patients whose CT scans clearly showed signs of viral infection but whose NATs tested came back negative were "released" back into their communities due to a shortage of hospital beds.

CT Scans

Okay so Ritz, this is all very informative what does this have to do with AI? Well we mentioned earlier about CT scans which are still most efficient way for detecting and evaluating the severity of pneumonia, according to AISC [3]. However, for doctors, providing the prognosis from the CT scan, is not a trivial task and can take some time to conclude. This is where AI comes in! AI assisted diagnosis could greatly reduce radiologist workload by providing:

• Early diagnosis,
• Early isolation, and
• Thus reduced virus spreading

There is a dataset that has over 46000 CT images from 106 patients over 26000 from COVID19 patients and over 20 000 images from 55 control group [4].

They use neural network called UNet++, where the U is because of the shape of the network [5]. So, Per patient it had an accuracy of 95.24% and per image there was an accuracy of 98.85%. While this is good, it is not the holy grail of automated diagnosis for Corona, but this is rather used to assist the radiologist and for quicker diagnosis. AI assisted diagnosis decreased reading time by 65%. So it has great potential to improve efficient of diagnosis and relieve pressure of radiologist in practice and relieve the already overworked and exhausted doctors.

Infer vision and Alibaba are both using computer vision on CT scan images for corona diagnosis. Alibaba has also come up with an AI system that is capable of detecting Corona virus in minimal time. They claim to have an accuracy of 96% for this system which is trained in 5000 confirmed test data. Reports say, at least 100 healthcare sector departments will be employing Alibaba’s AI system. As Per a report the from Nikkei’s Asian Review, Alibaba claims it only takes 20 seconds for the AI to make a determination – and according to the report, humans generally take about 15 minutes to diagnose the illness as there can be upwards of 300 images to evaluate. [6]

Looking at a few other related investigations, there is paper on Lung infection quantification in CT Images of COVID-19 using deep learning by Fei Shan et al. They achieved an average 91% average Dice similarity coefficient, which is also known as the overlap index, of 91.6% agreement between automatic and manual segmentations. So essentially this tells us how well the AI overlapped to human detections [7]. This process of theirs took 4 minutes, not as fast as Alibaba’s 20 seconds, but still good.

LinkingMed, released China’s first open-source AI model for pneumonia CT image analysis powered by Baidu’s open-source deep learning platform, PaddlePaddle. The By leveraging PaddlePaddle and the semantic segmentation toolkit PaddleSeg, LinkingMed has developed an AI-powered pneumonia screening and lesion detection system. The system can pinpoint the disease in less than one minute, with a detection accuracy of 92% and a recall rate of 97% on test data sets.

PyImageSearch posted on his blog of how you can use X-Ray image for detecting COVID-19 using Keras TensorFlow and Deep learning [8].

So, what we can conclude from this is that if you think you infected, its best to go for a CT Scan and AI can help speed up this process, thus alleviating the stress on doctors.

Tracking and Analytics for Preventative Measure

Why is it important to track? –Well, because when you get insights early on to pinpoint an event before they appear, we are able to proactively remedy the situation. So, in this case, if you are able understand the spread in real-time and predict disease outbreaks, you can prepare governments and health workers for the arrival of the virus and to take special measure or precautions such as quarantining potentially infected people. Data can be mine from a wide variety of sources such as, online blogs, chatrooms, google searches, twitter and other social platforms. So essentially AI can find patterns in these data sources and can help provide insights into emerging infectious diseases like the dominant Corona virus.

The biggest platforms that are currently tracking COVID-19 are

• John Hopkins Dashboard – Which monitor Confirmed cases vs recovered
• NextStrain – Which tracks how the virus spreads, and the most prominent is
• BlueDot – whom predicts infectious diseases, amongst others.

BlueDot

Focusing particularly on BlueDot, they are an AI-based startup who have built intelligent systems about people to determine the likelihood of occurrence of any disease.

• So, their AI platform is said to be amongst the latest technological advances using data analytics to map and prevent diseases.
• BlueDot predicted the SARS outbreak which turned out to be true.
• They also use Natural Language Processing (NLP) as a tool to track diseases from across the globe by analyzing different human languages and use the information through them.

GIS Technology

Geographic Information Systems (GIS) Technology is one important tool used in recognizing and stopping the spread of corona virus.

• This system has a data analytics dashboard that depicts all the case studies of Corona virus across the world.
• It uses the information to detect places where people talk and discuss the disease. Hence, making its data gathering critical.
• Social media is the best platform for data mining in this case where people are continuously talking about their interests and views towards such problems.
• In accordance with the situation, preventive measures can be drawn with analytics to track the spread of disease.

There is obviously other active research from companies like Google, Facebook in collaboration with Harvard School of Public Health, the NHI, WHO, CDC [Put their logos here] (Center for disease, control and prevention) amongst many others

If you would like to challenge yourself with the world’s most pressing problem using AI and data science, then I suggest you check out Kaggle. They have just announced that they’ve just publish COVID-19 Open Research Dataset or (CORD-19). CORD-19 is a resource of over 29,000 scholarly articles, including over 13,000 with full text, research about COVID-19, SARS-CoV-2, and related corona viruses. So, check that out if you would like to help out! [9]

So, summing up the tracking and analytics section, is that we can use AI and big data to help predict and track outbreaks around the world, and this helps us to take action and be prepared before an actual outbreak occurs to control the spread of the disease.

Reduction – Disease Control

Once AI has helped track and predict cases of Corona, its now time to figure out how to physically control the spread of the disease. Quarantining is still by far the best way to isolate branches of infected people.

Robots for Disease Control

Several Chinese companies are developing solutions to fight the deadly spread of Corona virus. The Chinese government is using robots along with AI to scan their surrounds and disinfect rooms in hospitals [10]. The Thor-1 robot looks like an artillery drone, but its ammunition is disinfectant liquid [11]. At airports, metro stations and other crowded areas, there are also patrolling robots and drones that use infrared cameras as a fever screening system. If the detected persons temperature, is above normal, it will raise an alert to a potential corona victim.

They would probably be using a person object detector for this.

Robots are also used to deliver food and medicines to avoid human to human contact [11]. Also, through a speaker, they will tell people to practice social distancing and wash their hand properly with soap for 20 seconds or use disinfectant.

Many Chinese companies have developed automated systems for contactless delivery, contactless ordering, and spraying disinfectants. Like Shenzhen-based Pudu Technology who have installed its robots for catering in more than 40 Chinese hospitals for this purpose.

Another Shenzhen-based company called MicroMultiCopter has supplied for drones in order to transport medical samples and conduct thermal imaging for identifying human congregation much more efficiently. Once identified, the drone alerts ground staff to take measure to disperse the congregation [12].

So the core of disease control of COVID-19 is to quarantine infected people and to promote social distancing. AI is used in robots and drones that assist with patrolling, screening, compliance and to carry out non-contact tasks 

Cure / Vaccine

Let’s look at other companies working towards potential treatments.

Exscientia

Exscientia’s CEO, Prof. Andrew Hopkins states that AI could be used in three ways in the current crisis, [13]:

• to rapidly develop antibodies and vaccines for the COVID-19 virus,
• to scan through existing drugs to see if any could be re-purposed,
• to design a drug to fight both the current and future corona virus outbreaks.

While this is cool, he also warned that people should be realistic about what AI could achieve.

The appeal of AI-designed drugs is relatively straightforward. There are literally tons of possible molecules that might be useful in medications, far too many for all the medical researchers in the world to manually test. But by using different types of AI, a computer system can simulate and mine through a wide variety molecules, comparing them against different parameters and synthesizing the most promising compounds faster than any human could. [14]

If quantum computers are capable enough, then cures and vaccines could arrive quicker even before a virus can become a pandemic.

Exscientia became the first company to put an AI-designed drug molecule to human trials earlier this year.

Insilico Medicine

Insilico is a biotech startup focused on disease prevention. They are an Analytics and AI-centric company, that are currently working on developing a solution that will pre-inform doctors about the fighter molecules capable of inhibiting the corona virus.

Dr. Zhavoronkov from Insilico says that they use GAN’s or Generative Adversarial Networks to create molecules that bind to proteins. He further states that “Over the course of four days, his company’s AI algorithms created 100,000 unique molecules which were narrowed down to 100 promising specimens. And from those 100, scientists chose 7 of those to synthesize. Reinforcement learning is then used to improve its conclusions or predictions through a feedback loop by guiding the molecules to specific objectives, such as ensuring that those molecules can still be effective after human metabolization. [15]

They hope to have potential treatments that they can begin testing in partnership with a pharmaceutical company in April, as reported by Wall Street Journal. Other biotech companies that are using AI on a molecular level are Vir Biotechnology, XtalPi and Atomwise Inc. [15]

Michigan State University

Technology could also mine through scientific research and patient data and help inspire the repurposing of old drugs, among a wide variety of other applications. And the uses can be even broader than just medication: [14]

Michigan State University (MSU) is using Machine Learning for new drug candidates for possible treatment of the corona disease. They state that the sequence identity of COVID-19 sequence identity is 96.1% to the SARS-COV. They also mentioned that they are in the process of analyzing the effectiveness of antiretroviral medication or anti-HIV/AID drugs and SARS-COV treatments for COVID-19 – No significant results have been reported to date as yet. [16]

Google’s DeepMind

Google’s UK team Deepmind, known mostly for its algorithms that beat human players at Go, DOTA, StarCraft and other games, has been been working directly on solutions for drug discovery. Their secret weapon? AlphaFold, which is a deep learning system that tries to predict protein structures accurately when no such proteins exist. With the genomic blueprint of COVID-19 being openly accessible deep mind is focusing on a few understudied but potentially important proteins that could hypothetically become drug or vaccine targets by using AI and machine learning. Their findings hope to cut down on months of effort typically used to determine the protein structure of a virus. [17]

Gilead Sciences, in collaboration with a Beijing hospital, has already started Phase III, human testing of an experimental antiviral drug, remdesivir, in Wuhan.

Global health authorities deem it the most promising of possible treatments for people who are severely ill with the novel corona virus. Because it is a “broad spectrum’’ drug that has been effective against multiple viral targets in the lab and in animals, the strategy could work, stated y experts. [18]

The patients will be intravenously administered with remdesivir, and the durations for two doses will be studied. The hope is that remdesivir will show better results with the corona virus than it did with Ebola, which is from a different viral family. The first trial results could only be available in April. [18]

Misinformation / Education

When combating a pandemic such as the corona virus, there is another thing that goes viral simultaneously. What you may ask? Well its fake news, also known as an Infodemic. A lot of people who do not have any background in medicine, biotechnology amongst others are conjuring their own treatments and so called “cures”. This is a serious problem, because people who are infected or are looking for preventative measures end up harming themselves even more because of this misinformation.

So how do we combat this?

We’ll for the well-being of you and you family. Please Follow only credible sources like the World health organization (WHO), the NHS and trusted organizations as well as public health agencies.

So far, there has not been any comprehensive study of how much misinformation remains on platforms such as Google and Facebook but it is likely to be substantial.

Google said its team was "working round the clock to safeguard our users from phishing, conspiracy theories, malware and misinformation". If you Search for coronavirus or Covid-19, and an SOS Alert appears, alongside links to help and information about the virus.

YouTube and Facebook, meanwhile, is using its homepage to direct users to the World Health Organization and other groups, for education and information, while working to remove videos suggesting alternative cures as soon as they go live.

Signing out

Sho, that was quite a long article, hope you enjoyed it and are better informed on how AI can be used to fight this pandemic. Remember that all the sources that I have used and mentioned in this video can be found in the video description down below and on my blog on www.AugmentedStartups.com.

I hope you stay safe during this period, hopefully it’ll pass soon, and we can go back to our normal way of life. Make sure to wash your hands thoroughly and practice social distancing.

Learn More in AI

If you'd like to learn how to develop your own AI apps in Computer Vision (AI-CV) check out our training called Ultimate AI-CV Practitioners PRO. This course will teach you how to become a PRO AI-CV developer with: Object Detection, Instance Segmentation, Pose Estimation, Android and Raspberry Pi Development! Click the link or image below to attend the FREE Webinar. – Click Here

Bibliography

[1] Wikipedia, "Nucleic acid test," [Online]. Available: https://en.wikipedia.org/wiki/Nucleic_acid_test.

[2] Straits Times, "Corona Virus US firm develops new test that can confirm cases in 30 minutes," [Online]. Available: https://www.straitstimes.com/asia/east-asia/coronavirus-us-firm-develops-new-test-that-can-confirm-cases-in-30-minutes.

[3] AISC, "ML Papers Explained – A.I. Socratic Circles – AISC," [Online]. Available: https://www.youtube.com/watch?v=W2ZdaBNudAY.

[4] T. Ai, Z. Yang, H. Hou and e. al., "Correlation of Chest CT and RT-PCR Testing in Coronavirus Disease 2019 (COVID-19) in China: A Report of 1014 Cases," RSNA, 2020.

[5] Towards Data Science, "Biomedical Image Segmentation: UNet++," 2019. [Online]. Available: https://towardsdatascience.com/biomedical-image-segmentation-unet-991d075a3a4b.

[6] Nikkei Asian Review, "Alibaba says AI can identify coronavirus infections with 96% accuracy," 19 February 2020. [Online]. Available: https://asia.nikkei.com/Spotlight/Coronavirus/Alibaba-says-AI-can-identify-coronavirus-infections-with-96-accuracy.

[7] S. Fei, G. Yaozong, W. Jun and e. al., "Lung Infection Quantification of COVID-19 in CT Images with Deep Learning," Cornell University, 2020.

[8] A. Rosebrock, "Detecting COVID-19 in X-ray images with Keras, TensorFlow, and Deep Learning," 16 March 2020. [Online]. Available: https://www.pyimagesearch.com/2020/03/16/detecting-covid-19-in-x-ray-images-with-keras-tensorflow-and-deep-learning/?__s=ymifacmisiz5sfwjthim.

[9] Kaggle, "COVID-19 Open Research Dataset Challenge (CORD-19)," [Online]. Available: https://www.kaggle.com/allen-institute-for-ai/CORD-19-research-challenge.

[10] "China turns to AI, robots in coronavirus control," New China TV, 11 February 2020. [Online]. Available: https://www.youtube.com/watch?v=7YVzWgyMbl8.

[11] "Coronavirus: China steps up use of robots to fight Covid-19 epidemic," South China Morning Post, 27 February 2020. [Online]. Available: https://www.youtube.com/watch?v=lQMMTw1f43s.

[12] Bloomberg, "China Is Using Drones to Fight Coronavirus," Bloomberg, [Online]. Available: https://www.youtube.com/watch?v=GIfURSuivVY. [Accessed 18 March 2020].

[13] BBC News, "Coronavirus: How can AI help fight the pandemic?," BBC, 12 March 2020. [Online]. Available: https://www.bbc.com/news/technology-51851292.

[14] Vox, "Would you take a drug discovered by artificial intelligence?," Vox, 31 January 2020. [Online]. Available: https://www.vox.com/2020/1/31/21117102/artificial-intelligence-drug-discovery-exscientia.

[15] WSJ, "https://www.wsj.com/articles/biotech-companies-tap-ai-to-speed-path-to-coronavirus-treatments-11583451564," Wall Street Journal, 6 March 2020. [Online]. Available: https://www.wsj.com/articles/biotech-companies-tap-ai-to-speed-path-to-coronavirus-treatments-11583451564.

[16] G. Kaifu, N. D. Duc and e. al., "Machine intelligence design of 2019-nCoV drugs," bioRxiv, p. 16, 2020.

[17] Singularity Hub, "DeepMind’s Protein Folding AI Is Going After Coronavirus," Singularity Hub, 17 March 2020. [Online]. Available: https://singularityhub.com/2020/03/17/how-deepminds-ai-is-working-to-decode-coronavirus/.

[18] Washington Post, "The best hope for coronavirus treatment is an experimental drug that fizzled against Ebola," Washington Post, 11 March 2020. [Online]. Available: https://www.washingtonpost.com/business/economy/the-best-hope-for-coronavirus-treatment-is-an-experimental-drug-that-fizzled-against-ebola/2020/03/10/8a9e8cd4-5fe8-11ea-b29b-9db42f7803a7_story.html.