Artificial Intelligence against the coronavirus: How Technologies Help in Fighting against COVID-19 | SBERMED AI
About company
Our solutions
Resource center

Contact us to find out more and start working with us





    * Required field

    SberMedAI LLC uses Cookies (files with information about past visits to the website) for service personalization and user convenience. To continue browsing the website, you need to enable cookies and accept our Cookies policy. You may disable cookies, but the website will become unavailable. Cookies policy

    Accept

    Artificial Intelligence against the coronavirus: How Technologies Help in Fighting against COVID-19

    January 31 2022

    |

    Reading time 11 minutes

    The new coronavirus posed social, political and economic challenges for mankind. On 30 January 2020 the World Health Organization (WHO)  declared that the outbreak constitutes a Public Health Emergency of International Concern (PHEIC).

    To ensure quick response,  researchers from various countries have to search for new ways of fighting against the pandemic. Artificial intelligence (AI) technology is a promising strategy for reaching objectives in health care and public organizations.

     

    AI helps in fighting against the pandemic

    AI is based on systems which can process data and make decisions on their basis. They are to organize work with large amounts of information automatically without human intervention. Artificial intelligence branches — machine learning and natural language processing algorithms. The latter is related to recognition of oral and written human speech. Due to this, machines can communicate with people by translating and generating messages.

    A smart system based on machine learning allows the following:

    uniting information and observations from various sources
    select data related to a research object
    • search for patterns and classify the data
    • make predictions based on the obtained model.

    AI is used for monitoring and predicting the epidemic, and for developing methods of prevention and treatment of the virus. As new data are obtained, artificial intelligence learns, and decisions are made more accurately.

     

    Monitoring COVID-19 outbreaks: artificial intelligence expands the search zone

    In traditional epidemiology, working with structured data for monitoring outbreaks is preferred.   Research is focused on official reports on the spread of a disease:

    place where an outbreak occurred
    period of time
    • number of new cases and deaths.

    Unstructured data enable considerable expansion of this list.  AI helps filtering and arranging them: removing noisy data and select meaningful information — alarm signals indicating a new wave of infections.

    Research can include:

    news in various languages
    information on airline ticket sales
    • flight maps
    • climate data
    • messages on medical forums and blogs about unusual cases of disease
    • information on cases of disease among animals
     examination of individuals, videos and images.

    A machine learning model has become a basis for BlueDot Insights platform by the Canadian AI company BlueDot. AI analyzes and processes not only official statistical reports, but also unstructured data. BlueDot Insights warns about possible outbreaks of COVID-19 and its ways of spreading among population close to real time.

    The platform was proved effective at the beginning of the pandemic. BlueDot Insights predicted the coronavirus outbreak in Wuhan on 31 December 2019. AI warned about the threat six days earlier than Centers for Disease Control and Prevention (CDC) and nine days earlier than WHO.

    Smart analysis of data is also implemented in HealthMap. The technology was developed by researchers from Harvard Medical School in cooperation with Boston Children’s Hospital.

    HealthMap scans local data to find information indicating COVID-19 outbreaks:

    • social media and blog posts
    messages in chats
    • web queries
    • press news

    Expected coronavirus outbreaks are displayed on a freely available map as colored dots.  HealthMap updates infographics as information is collected.

    AI-based on machine learning algorithms is a versatile tool for specialists.  Epidemiologists and virologists assess accuracy of predictions. Public health officials and and state authorities use data to fight against the epidemic and ensure quick response to possible outbreaks.

     

    Detect and prevent: how computer vision and neural networks identify signs of COVID-19

    AI-based temperature control systems identify fever, which is one of the main symptoms of the new coronavirus.  This helps isolate a person in a timely manner and prevent transmission of the virus. The Chinese company Megvii presented such a technology. It includes computer vision elements:

    body recognition
    face recognition
    • dual sensor scanning using infrared cameras and visible spectrum.

    Advantages of AI by Megvii include keeping a distance and continuous scanning. These advantages are useful in places with heavy passenger traffic: in airports, at metro stations and bus stops. Besides, no physical contact is necessary to take temperature. According to the company, temperature is taken at the speed of three to five persons per second, with precision of ± 0.3 ℃ (0.5 ℉).

    In February 2020, Alibaba DAMO Academy research group created AI for identifying COVID-19 on CT of the chest. Developers state that the process is quick and accurate: the system can detect signs of the coronavirus in 20 seconds with precision of 96%. DAMO technology helps assess suspicious cases and monitor dynamics of a confirmed disease.

    In Russia, SberMedAI team is implementing AI Resp project, which makes preliminary online diagnostics of the coronavirus possible. The service is researching a breathing pattern which is different when a person is healthy, has COVID-19 or suffers from another respiratory disease.

    Algorithm of the AI-based AI Resp mobile app:

    analyzing an audio with a patient’s breathing and speech
    comparing input acoustic characteristics with learning data – breathing, coughing and shortness of breath characteristic of COVID-19
    assessing COVID-19 probability based on obtained data.

    AI-based lung diagnostics is another SberMedAI’s area. The CT Lungs service recognizes respiratory symptoms of an infection and signs of respiratory cancer:

    identifies location of affected tissues
    identified degree and percentage of lung damage
    • identified minimal nodules which are larger than 4 mm.

    A radiologist has to perform many routine actions. AI by Russian developers makes diagnostics simpler and reduces potential errors when medical decisions are made. The CT Lungs service allows the following:

    selecting suspicious areas on chest CT images
    making necessary calculations for pathological areas
     prepare a complete opinion.

    Options of the CT Lungs service available to a health professional save time and resources for tackling other clinical tasks.

     

    In search for treatment: using AI in laboratories and clinics

    During the pandemic, off-label use of drugs with the help of AI allows “filling gaps” in therapy for patients with COVID-19.

    The British startup BenevolentAI considered the possibility of using baricitinib (which is used is to treat rheumatoid arthritis) for treating COVID-19. Using machine learning, AI processed large amounts of scientific and medical data. Due to this, potential antiviral properties of baricitinib were identified. The substance reduces binding of the virus to susceptible cells and suppresses disease development.

    EXAM clinical trial predicted the need of patients with COVID-19 for additional oxygen.  Specialists from Europe, Asia and the Americas took part in the project. To analyze information from all fur continents, federated machine learning was used. The model was taught using data from electronic medical records and chest X-ray images.  The algorithm ensures patient anonymity.

    The developed AI was tested in three independent centers, and results were good. After an emergency admission EXAM predicted the need for oxygen within 24 hours. Sensitivity and specificity were 95% and 88.2% respectively.

     

    Technologies suggest a vaccine formula

    many probable protein configurations, which include a true one
    a labor-intensive and expensive experiment stage
    • data processing which takes a long time
    • research which is conducted when the virus is highly variable and new variants form

    AI computing capabilities help science. Scientists from Baidu AI published Linearfold algorithm which predicts RNA secondary structures. Due to computing, total time of analyzing genomic data becomes 120 times shorter: from 55 minutes to 27 seconds.

    AlphaFold system made it possible to make a huge step forward in understanding protein structure of the virus. This is the project of the British company DeepMind which has been part of Google since 2014. Spike protein structure using AlphaFold is predicted in “free modeling” conditions, when a researcher faces a lack of information on structure of similar proteins.

    DeepMind considers its algorithm to be a computing platform for COVID-19 prevention and treatment. Priority areas include vaccine development and generating hypotheses to obtain a drug in experimental conditions.

     

    “Post COVID-19 condition” in IT: what comes next?

    During the crisis, new technologies made analytics much quicker at all levels: from a patient’s needs to requests to state authorities and public organizations.

    AI enables response to the spread of the virus, retaining a major part of resources in health care:

    predicting a new wave of infections
    screening for the main symptoms of the virus
     accelerating development of drugs and vaccine
     providing smart support for health professionals due to automating diagnostics and case monitoring

    Ethical issues of AI models are related to the following:

    processing large amounts of data Part of information may be confidential, and its use affects civil liberties.
    low level of transparency in AI systems Smart systems use complex data and algorithms. Therefore, it is difficult to assess their ethical aspects. Information in AI may be biased and not reflect interests of vulnerable members of society. A lack of independent supervision in centralized decision-making weakens public confidence.
    inadequate risk assessment Crisis scenarios require prompt solutions for saving lives. Therefore, benefits and risks of using AI cannot always be assessed.

    Conditions and limits of AI use require further discussion. Protecting individual interests in processing information, preparing freely available reports and performing adequate risk assessment are possible measures for enhancing public confidence in the future.

     

    Sources:

    1. Muzammil Khan, Muhammad Taqi Mehran et al. Applications of Artificial Intelligence in COVID-19 Pandemic: a Comprehensive Review [Electronic resource] // Elsevier Public Health Emergency. 2020. 185. URL: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8359727/ (access date: 24/12/2021). DOI: 10.1016 / j.eswa.2021.115695.

    2. Expert.ai. team What is Machine Learning?  A definition. // Expert.ai. 2020. URL: https://www.expert.ai/blog/machine-learning-definition/  (access date: 24/12/2021).

    3. Artificial Intelligence — What it is and why it matters // SAS Institute Inc. URL: https://www.sas.com/en_us/insights/analytics/what-is-artificial-intelligence.html (access date:: 24/12/2021).

    4. Vandana Janeja. Predicting the Coronavirus Outbreak: How AI Connects the Dots to Warn about Disease Threats [Electronic resource] // The Conversation 2020. URL: https://theconversation.com/predicting-the-coronavirus-outbreak-how-ai-connects-the-dots-to-warn-about-disease-threats-130772/ (access date:: 24/12/2021).

    4. Willies Ogola. How AI Predicted the Coronavirus Outbreak [Electronic resource] // Section. 2021. URL: https://www.section.io/engineering-education/how-ai-predicted-the-coronavirus-outbreak/ (access date:: 24/12/2021).

    6. BlueDot for Healthcare [Electronic resource] // Bluedot. URL: https://bluedot.global/healthcare/ (дата обращения: 24.12.2021).

    7. Virginia L. Ma. Harvard Medical School Professor’s Artificial Intelligence System Maps Coronavirus Outbreak [Electronic resource] // The Harvard Crimson. 2020. URL: https://www.thecrimson.com/article/2020/3/3/hms-coronavirus-tracking-tool/  (access date:: 24/12/2021).

    8. Adrian Cho. Artificial Intelligence Systems Aim to Sniff out Signs of COVID-19 Outbreaks [Electronic resource] // Science. 2020. URL: https://www.science.org/content/article/artificial-intelligence-systems-aim-sniff-out-signs-covid-19-outbreaks/  (access date:: 24/12/2021).

    9. How BlueDot Empowered Taiwan’s World-class COVID-19 Response [Electronic resource] // BlueDot. URL: https://bluedot.global/clients/taiwan/  (access date:: 24/12/2021).

    10. What is Computer Vision? [Electronic resource] // IBM. URL: https://www.ibm.com/topics/computer-vision/ (access date:: 24/12/2021).

    11. MEGVII’s Ming Ji Mini Solution Deployed in global markets to Combat COVID-19 [Electronic resource] // Megvii Technology Limited. 2020. URL: https://en.megvii.com/news_detail/id/165 (access date:: 24/12/2021).

    12. Cecilia Li. How DAMO Academy’s AI System Detects Coronavirus Cases [Electronic resource] // Alizila. URL: https://www.alizila.com/how-damo-academys-ai-system-detects-coronavirus-cases/ (access date:: 24/12/2021).

    13. Peter Richardson, Ivan Griffin, Catherine Tucker et al. Baricitinib as Potential Treatment for 2019-nCoV Acute Respiratory Disease [Electronic resource] // The Lancet. 2020. URL: https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(20)30304-4/fulltext?rss=yes#%20 (access date:: 24/12/2021).

    14. Craig Brierley. World First for AI and Machine Learning to Treat COVID-19 Patients Worldwide [Electronic resource] // University of Cambridge. 2021. URL: https://www.cam.ac.uk/research/news/world-first-for-ai-and-machine-learning-to-treat-covid-19-patients-worldwide  (access date:: 24/12/2021).

    15. Ittai Dayan, Holger R. Roth et al. Federated Learning for Predicting Clinical Outcomes in Patients with COVID-19 [Electronic resource] // Nature Medicine. 2021. URL: https://www.nature.com/articles/s41591-021-01506-3 (дата обращения: 24.12.2021). DOI: 10.1038/s41591-021-01506-3 .

    16. Baidu. How Baidu is Bringing AI to the Fight against Coronavirus  [Electronic resource] // MIT Technology Review. 2020. URL: https://www.technologyreview.com/2020/03/11/905366/how-baidu-is-bringing-ai-to-the-fight-against-coronavirus/  (access date:: 24/12/2021).

    17. Computational Predictions of Protein Structures Associated with COVID-19 [Electronic resource] // DeepMind. 2020. URL: https://deepmind.com/research/open-source/computational-predictions-of-protein-structures-associated-with-COVID-19 (access date:: 24/12/2021).

    18. Tzachor, A., Whittlestone, J., Sundaram, L. et al. Artificial Intelligence in a Crisis Needs Ethics with Urgency [Electronic resource] // Nature Machine Intelligence. 2, 365–366 (2020). URL: https://www.nature.com/articles/s42256-020-0195-0#citeas  (access date:: 24/12/2021). DOI: 1038/s42256-020-0195-0.

    19. Stephen Cave, Jess Whittlestone, Rune Nyrup et al. Using AI Ethically to Tackle Covid-19 [Electronic resource] // BMJ. 2021; 372: n364. URL: https://www.bmj.com/content/372/bmj.n364 (access date:: 24/12/2021). DOI: 10.1136/bmj.n364.

    Похожие статьи