All Eyes on COVID-19 and AI: The CoronaHack
Andra Stefan
As of today, the novel coronavirus has infected 7,495,164 people and killed 421,976 according to WHO. [1] Moreover, as a result of the lockdown measures taken globally to contain the spread of the virus, our day-to-day lives have changed dramatically and the world economy has been disrupted. Given that Artificial Intelligence (AI) research in medicine has grown exponentially in the recent years, it is unsurprising that AI has stepped in to fight the COVID-19 pandemic.[2] AI technologies can address almost every aspect of the disease from prevention and diagnosis to prognosis and treatment. In this article, I present some of the most promising AI-powered projects that aim to tackle this global health crisis.
Prevention
The AI-based early warning system developed by the Canadian firm BlueDot was among the first in the world to accurately predict the risk of spread of the coronavirus. BlueDot’s clients, ranging from governments to hospitals and airlines, were all warned of the emerging risk from COVID-19 back in late December 2019.[3] Thanks to BlueDot’s software, the Californian government was able to take appropriate measures promptly, closing the state borders ahead of any other state in the US.[4] In order to anticipate the impact of disease locally and globally, this warning system utilizes a wide range of datasets including flight itineraries, demographics, climate conditions and health system capacities. [3] On the other hand, the machine learning epidemic model developed by a research group from UCLA can predict the number of confirmed COVID-19 cases, death counts and recovered cases. This model is helping public health institutes in the US predict the evolution of the pandemic and aids in evaluating the effectiveness of governmental policies. What distinguishes this model from others is that it does not rely only on observed data, namely the reported number of cases and fatalities. Instead, it infers the number of untested and unreported COVID-19 cases and utilizes these inferred data to enhance the predictive accuracy of the model.[5]
Diagnosis
Real-time PCR (RT-PCR) is the gold standard test for the diagnosis of COVID-19, however, it can take up to two days to obtain the results and is associated with a high risk of eliciting false-negatives. [6] On the other hand, lung Computed Tomography (CT) was shown to have a higher sensitivity than RT-PCR in detecting COVID-19. As a result, AI algorithms have been developed to assist radiologists in diagnosing COVID-19 based on CT scans. For instance, the icolung software accurately identifies lung lesions caused by coronavirus on CT scans much faster than experienced radiologists. This software is the first AI solution for COVID-19 to receive CE-mark (certification mark) and is currently available for clinical use. [7] Moreover, AI diagnostic algorithms can predict whether a person is infected with the novel coronavirus even without testing. An AI model was shown to achieve 80% diagnostic accuracy in predicting COVID-19 based on age, gender and a set of symptoms. This algorithm will soon undergo clinical trials in the UK and the US.[8]
Prognosis
AI tools can also help triaging patients which is key to controlling the spread of the virus in healthcare facilities and managing cases effectively. For instance, researchers from Maastricht University have developed an AI algorithm that assesses the risk of severe disease requiring mechanical ventilation based on available patient data.[9] Besides risk stratification, an AI platform developed by the Israeli start-up Diagnostic Robotics also enables personalised guidance based on patient self-assessment as well as remote monitoring of sick patients.[10]
Treatment
AI technologies are accelerating the search for an effective treatment for COVID-19. The UK-based company BenevolentAI utilized their AI-powered drug discovery platform to mine scientific literature in order to identify potential targets and corresponding therapies against SARS-CoV-2. Consequently, BenevolentAI identified a class of approved drugs that inhibit cellular endocytosis thereby preventing viral entry into the cells. Out of these drugs, Baricitinib – a drug currently approved for rheumatoid arthritis – appeared to be the most effective and is already in late-stage clinical trials in the US. [11] Another UK biotech company, Exscientia, has established a partnership with the California Institute for Biomedical Research (Calibr) and Diamond Light Source to search for potential COVID-19 therapies. Exscientia has thus gained access to Calibr’s 15 000 drugs that have passed clinical safety tests and will screen this large collection of molecules using AI to identify those that target key viral proteins. This work is supported by the Diamond Light Source’s crystallography facility which enables the determination of the 3D structure of viral proteins.[12]
It is truly inspiring to see how promptly many organisations have mobilised to contribute to the battle against COVID-19 by developing AI solutions. Overall, the key strengths of AI technologies are their high predictive accuracy and the rapidity with which they process information and return the results that are needed. However, we must bear mind that AI is only a tool and that the human input, be it in terms of large scale or smaller scale decision-making, is still key to delivering an appropriate response to the current crisis.
REFERENCES
[1] World Health Organization. WHO Coronavirus Disease (COVID-19) Dashboard [Internet]. 2020 [cited 31.05.2020] Available from:
[2] Amisha, Malik P, Pathania M, Rathaur VK. Overview of artificial intelligence in medicine. J Family Med Prim Care. 2019;8(7):2328‐2331
[3] BlueDot. BlueDot Explorer [Internet]. [cited 31.05.2020] Available from: https://bluedot.global/products/explorer/
[4] Whitaker B. The computer algorithm was among the first to detect the coronavirus outbreak [Internet]. 2020. [cited 31.05.2020] Available from: https://www.cbsnews.com/news/coronavirus-outbreak-computer-algorithm-artificial-intelligence/
[5] UCLA Samueli Newsroom. UCLA Machine-Learning Model is helping CDC predict the spread of COVID-19 [Internet]. 2020. [cited 31.05.2020] Available from: https://samueli.ucla.edu/ucla-machine-learning-model-is-helping-cdc-predict-spread-of-covid-19/
[6] Tahamtan A, Ardebili A. Real-time RT-PCR in COVID-19 detection: issues affecting the results. Expert Rev Mol Diagn. 2020; 20(5):453‐454.
[7] Smeets D. The role of imaging AI and CT in COVID-19 [Internet]. 2020. [cited 31.05.2020] Available from: https://icometrix.com/resources/The-role-of-imaging-AI-and-CT-in-COVID-19
[8] ScienceDaily. New AI diagnostic can predict COVID-19 without testing [Internet]. 2020 [cited 31.05.2020] Available from: https://www.sciencedaily.com/releases/2020/05/200511112628.htm
[9] Maastricht University. Artificial Intelligence and the combat of the coronavirus [Internet]. 2020 [cited 31.05.2020] Available from: https://www.maastrichtuniversity.nl/news/artificial-intelligence-and-combat-coronavirus
[10] HospiMedica. AI-based triage and monitoring system predicts spread of coronavirus [Internet]. 2020. [cited 31.05.2020] Available from: https://www.hospimedica.com/covid-19/articles/294781936/ai-based-triage-and-monitoring-system-predicts-spread-of-coronavirus.html
[11] Cookson C. Biotechs harness AI in battle against COVID-19 [Internet]. 2020. [cited 31.05.2020] Available from: https://www.ft.com/content/877b8752-6847-11ea-a6ac-9122541af204
[12] Rees V. AI technology to screen existing drugs for use against COVID-19 [Internet]. [cited 31.05.2020] Available from: https://www.drugtargetreview.com/news/59188/ai-technology-to-screen-existing-drugs-for-use-against-covid-19/
Comments
Post a Comment