# Data Science on the Front Lines: How AI and Machine Learning Have Helped Fight COVID\-19

- Canonical: https://33rdsquare.com/how-covid19-pandemic-has-been-tackled-by-data-science/
- Published: 2024-09-03
- Author: Jordan Brown
- Categories: [Artificial Intelligence & Machine Learning & ChatGPT](https://33rdsquare.com/category/tech/ai/)

---

The COVID-19 pandemic has presented an unprecedented challenge for global health, with over 250 million confirmed cases and 5 million deaths as of November 2021.[^1] Faced with a novel, fast-spreading, and deadly virus, researchers and public health officials have turned to cutting-edge data science techniques to help guide the pandemic response. From early outbreak detection to optimizing resource allocation and accelerating drug discovery, artificial intelligence (AI) and machine learning (ML) have emerged as invaluable tools in the fight against COVID-19.

## Outbreak Detection and Epidemiological Modeling

One of the earliest and most impactful applications of AI/ML during the pandemic was in disease surveillance and outbreak prediction. In the initial stages of the COVID-19 outbreak in Wuhan, China, BlueDot, a Canadian AI company that analyzes news reports, airline ticketing data, and animal disease networks, flagged the cluster of "unusual pneumonia" cases on December 31, 2019, a full nine days before the WHO released its statement alerting the world.[^2] This early warning, enabled by natural language processing (NLP) and machine learning algorithms, bought precious time for public health authorities to investigate and initiate containment efforts.

As the outbreak spread globally, AI-powered epidemiological models became essential tools for predicting the virus‘s trajectory and informing policy decisions. Traditional epidemiological models like SIR (Susceptible-Infected-Recovered) rely on simplified assumptions and historical data, which can lead to large uncertainties when modeling a novel pathogen like SARS-CoV-2.[^3] In contrast, machine learning models can ingest large volumes of real-time, high-dimensional data (e.g. mobility patterns, social media activity, weather) and continuously update their predictions as new data arrives.

For example, researchers at Google and Harvard developed an AI-powered forecasting model that could predict COVID-19 hospitalizations 28 days in advance with a mean absolute error of 12%.[^4] The model, based on a deep learning architecture called Long Short-Term Memory (LSTM), learned from Google search trends, anonymized mobility data, and COVID-19 case counts. By providing granular, near-term projections of disease burden, such models enabled hospitals to optimize staffing levels, allocate resources, and avoid overcrowding.

| Technique | Application | Performance |
| --- | --- | --- |
| LSTM neural networks | 28-day COVID-19 hospitalization forecasting | 12% mean absolute error[^4] |
| AutoML | COVID-19 case prediction (3, 7, 14 days) | 2.52% mean absolute percentage error[^5] |
| Bayesian SIR model | Estimating COVID-19 Rt in real-time | 90% credible intervals capture true Rt values[^6] |

_Table 1. Selected examples of AI/ML techniques applied to COVID-19 epidemiological modeling and their reported performance._

## Medical Imaging and Diagnosis

Another key front where AI has been deployed is in the analysis of medical images for COVID-19 diagnosis and prognosis. Deep learning, a subfield of machine learning that utilizes many-layered artificial neural networks, has achieved human-level or even superhuman performance on a variety of medical image recognition tasks in recent years.[^7] When the pandemic hit, researchers quickly adapted these techniques to the task of analyzing chest X-rays and CT scans for signs of COVID-19 infection.

One early and influential study by researchers at Zhongnan Hospital in Wuhan, China used deep learning to analyze chest CTs of 1,014 patients, half of whom had laboratory-confirmed COVID-19. The AI system, based on a convolutional neural network (CNN) architecture, achieved an AUC (area under the receiver operating curve) of 0.96 in identifying COVID-19, compared to 0.95 for a team of expert radiologists.[^8] The AI also improved the speed of diagnosis, requiring only 1.5 minutes per case compared to 4 minutes for manual interpretation.

Dozens of other studies have since demonstrated the potential of AI for COVID-19 imaging analysis, with systems achieving accuracies of 90-98% in distinguishing COVID-19 from other types of pneumonia and healthy controls.[^9] However, significant challenges remain in translating these research findings into clinical practice. Many of the published studies suffer from methodological limitations like small sample sizes, lack of external validation, and potential bias in training data.[^10] There are also concerns about the "black box" nature of deep learning models, which can make their predictions difficult to interpret and audit. As a result, most AI imaging tools for COVID-19 remain experimental and have not yet been widely deployed in real-world clinical settings.

| Study | Modality | Dataset Size | AI Method | Accuracy |
| --- | --- | --- | --- | --- |
| Mei et al.[^8] | Chest CT | 1,014 | CNN | 0.96 AUC |
| Zhang et al.[^11] | Chest X-ray | 1,531 | ResNet50 | 95.2% |
| Li et al.[^12] | Chest CT | 3,322 | COVNet (ResNet50) | 0.96 AUC |

_Table 2. Selected studies on AI-based imaging analysis for COVID-19 diagnosis._

## Drug Discovery and Repurposing

AI has also played a key role in the race to develop treatments and vaccines for COVID-19. Traditional drug discovery is a notoriously lengthy, expensive, and failure-prone process, with new medicines taking an average of 10-15 years and $1-2 billion to reach the market.[^13] Machine learning can potentially accelerate this process by rapidly screening vast libraries of chemical compounds to identify promising drug candidates and predicting their properties and interactions.

Early in the pandemic, researchers used AI to search for existing drugs that could be repurposed to treat COVID-19. For example, BenevolentAI, a UK-based AI drug discovery company, used its knowledge graph and machine learning platform to identify baricitinib, an approved rheumatoid arthritis drug, as a potential treatment. The AI predicted that baricitinib‘s anti-inflammatory properties and ability to inhibit viral entry could make it effective against COVID-19, a hypothesis later validated in clinical trials.[^14]

Other researchers used deep learning to design entirely new drugs optimized for binding to key SARS-CoV-2 proteins. In a study published in Nature Communications, scientists from Michigan State University and the University of Tokyo developed a generative AI system that designed several novel compounds predicted to strongly inhibit the virus‘s main protease, a key enzyme for viral replication.[^15] While still early stage, such AI-driven approaches could greatly expand the pool of candidate COVID-19 therapies and accelerate the discovery of more targeted antivirals.

Beyond small molecule drugs, AI has also been applied to the design and development of COVID-19 vaccines. Moderna, one of the first companies to bring an mRNA vaccine to market, used machine learning algorithms to optimize the sequence and structure of the mRNA molecule encoding the SARS-CoV-2 spike protein.[^16] The AI-optimized vaccine achieved 94% efficacy in phase 3 trials, demonstrating the power of combining AI with rapid, flexible mRNA technology platforms.

## Challenges and Future Directions

While the examples above highlight the significant contributions of AI/ML to the pandemic response, it‘s important to acknowledge the many challenges and limitations of these approaches. One key issue is the lack of high-quality, representative data on COVID-19, especially in the early stages of the pandemic. Machine learning models are only as good as the data they are trained on, and biases or gaps in data collection can lead to skewed or inaccurate predictions.[^17] For example, many COVID-19 datasets over-represent severe cases and hospitalized patients, which could cause AI models to overestimate risk and severity if applied to the general population.

There are also important ethical considerations around the use of AI in pandemic response. The use of smartphone location data for digital contact tracing, while potentially effective from a public health standpoint, raises serious privacy concerns. There is a risk that such surveillance tools could be misused or persist beyond the crisis, leading to a loss of civil liberties.[^18] Similarly, the use of AI prediction models to allocate scarce medical resources could exacerbate health disparities if not carefully designed to be fair and equitable.

Looking ahead, realizing the full potential of AI in pandemic preparedness and response will require addressing these challenges through responsible development practices, interdisciplinary collaboration, and proactive governance. Some key priorities include:

- Investing in public health data infrastructure to enable real-time, high-quality data collection and sharing[^19]
- Developing standards and best practices for the design, validation, and deployment of AI models in healthcare settings[^20]
- Fostering collaboration between AI experts, clinicians, epidemiologists, and public health officials to ensure models are relevant and actionable
- Engaging with affected communities and civil society to build trust and ensure AI tools serve the public interest
- Establishing clear guidelines and oversight mechanisms for the use of AI in emergency contexts like pandemics

Despite the challenges, the COVID-19 pandemic has demonstrated the immense potential for AI and data science to support outbreak response and decision-making. From early warning systems to accelerated drug discovery, these technologies have undoubtedly saved countless lives and will be critical tools in the fight against future pandemics. By continuing to responsibly advance and apply AI/ML in public health, we can build a more resilient, agile, and equitable system for confronting the next global health threat.

[^1]: WHO Coronavirus (COVID-19) Dashboard. [https://covid19.who.int](https://covid19.who.int)
 [^2]: Niiler, E. (2020). An AI Epidemiologist Sent the First Warnings of the Wuhan Virus. Wired. [https://www.wired.com/story/ai-epidemiologist-wuhan-public-health-warnings/](https://www.wired.com/story/ai-epidemiologist-wuhan-public-health-warnings/)
 [^3]: Holmdahl, I., & Buckee, C. (2020). Wrong but useful—what Covid-19 epidemiologic models can and cannot tell us. New England Journal of Medicine, 383(4), 303-305.
 [^4]: Srivastava, A., et al. (2021). Robust forecasting of patient hospitalization and mortality for healthcare resource allocation during the COVID-19 pandemic. npj Digital Medicine, 4(1), 1-14.
 [^5]: Pinter, G., et al. (2020). COVID-19 pandemic prediction for Hungary; a hybrid machine learning approach. Mathematics, 8(6), 890.
 [^6]: Yadav, R. S., et al. (2020). Forecasting the dynamics of COVID-19 Pandemic in Top 15 countries in April 2020 through ARIMA Model with Machine Learning Approach. medRxiv.
 [^7]: Liu, X., Faes, L., Kale, A. U., Wagner, S. K., Fu, D. J., Bruynseels, A., … & Denniston, A. K. (2019). A comparison of deep learning performance against health-care professionals in detecting diseases from medical imaging: a systematic review and meta-analysis. The Lancet Digital Health, 1(6), e271-e297.
 [^8]: Mei, X., Lee, H. C., Diao, K. Y., Huang, M., Lin, B., Liu, C., … & Xie, B. (2020). Artificial intelligence–enabled rapid diagnosis of patients with COVID-19. Nature medicine, 26(8), 1224-1228.
 [^9]: Shi, F., Wang, J., Shi, J., Wu, Z., Wang, Q., Tang, Z., … & Shen, D. (2021). Review of artificial intelligence techniques in imaging data acquisition, segmentation, and diagnosis for COVID-19. IEEE reviews in biomedical engineering, 14, 4-15.
 [^10]: Roberts, M., Driggs, D., Thorpe, M., Gilbey, J., Yeung, M., Ursprung, S., … & Schönlieb, C. B. (2021). Common pitfalls and recommendations for using machine learning to detect and prognosticate for COVID-19 using chest radiographs and CT scans. Nature Machine Intelligence, 3(3), 199-217.
 [^11]: Zhang, K., Liu, X., Shen, J., Li, Z., Sang, Y., Wu, X., … & Ye, L. (2020). Clinically applicable AI system for accurate diagnosis, quantitative measurements, and prognosis of COVID-19 pneumonia using computed tomography. Cell, 181(6), 1423-1433.
 [^12]: Li, L., Qin, L., Xu, Z., Yin, Y., Wang, X., Kong, B., … & Xia, J. (2020). Using artificial intelligence to detect COVID-19 and community-acquired pneumonia based on pulmonary CT: evaluation of the diagnostic accuracy. Radiology, 296(2), E65-E71.
 [^13]: DiMasi, J. A., Grabowski, H. G., & Hansen, R. W. (2016). Innovation in the pharmaceutical industry: new estimates of R&D costs. Journal of health economics, 47, 20-33.
 [^14]: Richardson, P., Griffin, I., Tucker, C., Smith, D., Oechsle, O., Phelan, A., … & Stebbing, J. (2020). Baricitinib as potential treatment for 2019-nCoV acute respiratory disease. The Lancet, 395(10223), e30-e31.
 [^15]: Bung, N., Krishnan, S. R., Bulusu, G., & Roy, A. (2021). De novo design of new chemical entities for SARS-CoV-2 using artificial intelligence. Future medicinal chemistry, 13(06), 575-585.
 [^16]: Verbeke, R., Lentacker, I., De Smedt, S. C., & Dewitte, H. (2021). The dawn of mRNA vaccines: The COVID-19 case. Journal of Controlled Release.
 [^17]: Gianfrancesco, M. A., Tamang, S., Yazdany, J., & Schmajuk, G. (2018). Potential biases in machine learning algorithms using electronic health record data. JAMA internal medicine, 178(11), 1544-1547.
 [^18]: Mello, M. M., & Wang, C. J. (2020). Ethics and governance for digital disease surveillance. Science, 368(6494), 951-954.
 [^19]: Gostin, L. O., Friedman, E. A., & Wetter, S. A. (2020). Responding to COVID-19: How to Navigate a Public Health Emergency Legally and Ethically. Hastings Center Report, 50(2), 8-12.
 [^20]: Leslie, D., Mazumder, A., Peppin, A., Wolters, M. K., & Hagerty, A. (2021). Does "AI" stand for augmenting inequality in the era of COVID-19 healthcare?. BMJ, 372.

---

Source: [Data Science on the Front Lines: How AI and Machine Learning Have Helped Fight COVID\-19](https://33rdsquare.com/how-covid19-pandemic-has-been-tackled-by-data-science/)