Contact Us

This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.

Thank you for your interest in Cognizant.

We’ll be in touch soon.

We are sorry. Unable to submit your request.

Please try again or post your inquiry to

AI & data: our best hope for infectious disease prevention, treatment, mitigation & cures

May 14, 2020 - 299 views

AI & data: our best hope for infectious disease prevention, treatment, mitigation & cures

Advanced forms of AI are already critical to fighting the pandemic, and they'll be increasingly vital in combatting infectious disease in the future. 

Great challenges require great tools. And in a world where a virus such as SARS-CoV-2 has had a huge global impact in just a matter of months, its complexity and speed require new methods for finding solutions. Advanced forms of artificial intelligence (AI) are already playing a critical role in our pandemic-fighting efforts, and they’re expected to play an increasingly vital role in the war against infectious disease in the future

We see AI advancing our infectious disease fighting capabilities across the following dimensions:

  • Diagnosis: Patient evaluation – which includes reading imaging modalities such as X-rays – is now happening at a scale beyond the capacity of most medical staffs. To accelerate diagnosis and triage patient care, technologists and clinicians are training AI-based image recognition systems to identify signs of COVID-19. This includes the use of machine-learning (ML) programs that are trained to detect patterns in scanned lung images to understand who needs the most urgent care and access to limited resources.
  • Health: Given the wide-ranging impact of the virus across every demographic, we’ll need more advanced methods to understand the social determinants of any new disease and how it impacts transmission and recovery. Where people live, who they live with, how they feel, how they migrate and their daily habits are all important aspects for minimizing pandemic disease transmission. AI can turn individual-, cohort- and population-related data into actionable insights to optimize treatment regimens. With its superior processing capacity, AI can also more quickly and precisely recognize patterns and make associations than humans can.
  • Treatment: It takes years to commercialize effective vaccines. This work requires deep dives into complex virus structures to identify which proteins – of the vast numbers to evaluate – could impact how the resulting disease morphs and spreads. Using AI, researchers can develop models that simulate how a virus evolves, transmits, infects and propagates to find the best candidates for potential vaccines. This turns a universe of possibilities into a neighborhood we can explore.
  • Mitigation: AI can help us understand and predict where virus containment strategies are likely to be most effective. Some particulars include imaging systems that have been deployed to monitor social distance; thermal cameras with facial recognition capabilities that monitor employees’ health; and traffic analysis that monitors the possible spread among populations. Socioeconomic variables need to be considered, as well. Again, scale, complexity and urgency demand the use of AI.  

An Even More Powerful Form of AI: Evolutionary AI

Whatever contribution AI ultimately makes in the quest for a COVID-19 treatment or cure, machine intelligence is increasingly playing key roles throughout the global pharma R&D ecosystem, from molecule discovery through vaccine trial subject choices, as well as health predictions. Meanwhile, Evolutionary AI (an advanced form of machine intelligence) is poised to fulfill AI’s promise as an essential adjunct in analysis, comprehension and decision making for life sciences and healthcare organizations.

Evolutionary AI gives researchers a tool to dynamically model therapies, treatments and disease progression. This approach not only improves prediction accuracy but also informs and enhances decision making by offering a broader range of options (including probability of success, next-best options, etc.) and a more nuanced understanding of whatever is being studied. Because of these capabilities, Evolutionary AI promises models that can be built on-demand with greater precision using real-time data and fewer data scientists. Unlike current models that are either static or require constant tuning, which renders them less capable of contending with novel disease strains, Evolutionary AI models can be improved or adapted over time.

While data requirements render linear programming models and even deep-learning neural networks inadequate for modeling the outbreak of today’s novel coronavirus, Evolutionary AI applies algorithms to adapt models to changing conditions and objectives. By continuously generating the best decision strategy against a data-driven predictive surrogate, models can be evolved to increase accuracy and to adapt and improve over time as new data emerges.

Applying Evolutionary AI to Real-World COVID-19 Advancements

Within the dimensions defined above, numerous specific applications of Evolutionary AI are becoming available. Here are some key ones:

  • Accelerating and improving diagnosis: In addition to the vast span of available tests in the pipeline, a growing number of diagnostic tools are emerging that capitalize on AI’s powerful abilities. One recently developed AI tool has helped reveal which early symptoms (some surprising) may indicate the likelihood of severe illness ahead – a critical piece to solving the COVID-19 puzzle. We are pursuing a project that could accelerate the speed and accuracy of in-field COVID-19 diagnosis based on X-ray images. We evolved a deep network against 14 diseases, exploiting all their commonalities and differences. While we did not have sufficient data on each disease itself, because we were training on all 14 diseases simultaneously, we were able to exceed the state-of-the-art on detection, all from X-rays. After procuring a COVID-19 data set – roughly 150 X-rays indicating whether someone was positive or negative for the disease – we added the COVID-19 images as disease #15, and ran the AI analyses via our evolutionary deep network generation approach. We are now hitting 94% accuracy for COVID-19. Once verified against a larger dataset, these results are within acceptable deployment accuracy range. Furthermore, the deep networks can also be made small using multi-objective optimization of the architecture search process, to the point where they can be run on a laptop, making them viable for wide-scale field usage in diagnosing COVID-19.  When such deep networks are deployed and digitally connected, the cumulative data will contribute further to expanding and accelerating our ability to make initial diagnoses.
  • Gaining more insight into population health: In the fight against COVID-19 and in improving overall health, we know that the environment in which an individual lives and each person’s unique physiology exponentially increase the number of variables that need to inform our modeling capabilities to improve resultant forecasting capabilities. Macro data from the patient, along with socioeconomic data, provide a much more complete health picture. By applying AI to societal-level data, researchers, providers and authorities could assess and predict the risk and health of individuals, cohorts or populations, based on a multitude of environmental and social determinants. Worldwide cooperation among medical and science organizations – for communicating potential viral mutations, for example – has already been a major factor in managing this disease. Facilitated by wearables and other point-of-use medical/wellness technologies, health data and personal metadata is becoming more standardized and accessible for mining by Evolutionary AI for insights. One example is Kinsa’s internet-connected “smart” thermometer, which helps rapidly detect potential COVID-19 hot spots (in a non-seasonal-flu environment) via basal temperature increases. The unit gathers the source data, which is then processed by machine-learning technology that cleans, processes and makes sense of the data, allowing technicians to derive actionable insights. This is still a step or two behind Evolutionary AI, as the technology does not yet directly guide actions, but it shows how even simple physical-health data streams can be plied for powerful insights and results. A well-known example of AI’s vaunted benefits is BlueDot’s early flagging of COVID-19 as a potential pandemic by applying the technology to widespread travel and health data. However, that application just nicks the surface of how AI can improve public health and policy decisions on alleviating the potential health effects of pandemics. 
  • Quickly discovering effective treatments: The need for widespread testing is fundamental to our understanding of the disease – and when deployed at scale across the nation, the volume of data will challenge our ability to accurately  determine the prevalence and incidence of COVID-19. Insights gained from incidence and prevalence data will allow us to understand and determine the effectiveness of strategies such as social distancing – and how clinical interventions have benefited segments of the population. Only through AI can we rapidly translate all this information into relevant clinical insights that guide treatment of COVID-19 patients.  Baidu, one of the world’s largest AI and internet companies, is engaged in multiple COVID-19-related efforts, including sharing its LinearFold algorithm that has allowed the company to reduce viral RNA sequencing from about an hour to half a minute, a potent process upgrade in the search for a vaccine or treatment. AI startup BenevolentAI recently announced it had found a promising potential COVID-19 treatment – an existing treatment for rheumatoid arthritis. The drug, baricitinib, which has both antiviral and anti-inflammatory (specifically, anti-cytokine) properties, has already entered trials and was identified as potentially viable via AI meta-analysis that took three person-days and 90 minutes of computing time. Another treatment research track we are involved in concerns dealing with hospitalized patients’ nutritional needs and responses – how a patient’s specific condition during treatment or recovery varies with different nutritional programs. The applicable pharmacological interventions must be factored in as well, of course, to identify what is best for patient treatment and then to continuously readjust in the process.  A crucial phase of the vaccine or drug testing process is human clinical trials, with some COVID-19 candidates being tested on an accelerated basis. (Regulatory relief is a key dimension, and the industry has benefited from some relaxation in this regard.) In the clinical trial process, AI helps identify possible patient populations or trial participants by compressing time-to-market via simulations and analyses that point researchers to the best bets among potential targets for trial submission. For example, by viewing the error bars on all patients going through a clinical trial, AI can discern social determinant variations (not detectable by humans unaided by the augmentation of machine intelligence) that will help determine a given treatment’s potential safety and efficacy.
  • Mitigating the spread of the disease: AI will also make a difference in determining which nontherapeutic interventions should be in place, when they can be concluded and what is likely to happen thereafter. AI can calculate the impact of stringency measures and their relaxation, while accounting for a huge range of disease variables such as a given region’s stage in “flattening the curve,” how an area compares with other locations, the health and demographics of a given population, the behavior of previous pandemics and many more parameters. Pandemic mitigation efforts require societal collaboration and active participation. Public health and policy makers need access to reliable, timely information, and citizens need access to clear data in order to comply with medical guidance. Misinformation and disinformation are destructive to an adequate societal response. AI, via natural language processing (NLP), can prove invaluable in flagging unreliable content by creating text classification systems and otherwise determining the veracity of information sources.

The AI Day Has Dawned

While AI is not a panacea, we are learning how to wield it for the benefit of science, for the benefit of medicine and for the benefit of societal response and success in mitigating the effects of pandemics arising from infectious diseases.       Life sciences companies are increasing or maintaining their AI investments. Many new machine intelligence projects are brewing, and older ones are being remodeled. However, these investments and their potential public health benefits need to fit within an evolving societal debate on privacy and surveillance issues.

Despite privacy and bias concerns, AI remains a bright star in the life sciences firmament. A vaccine might yet be months or years away, if one is successfully developed at all. But AI is here and now, and in conjunction with unprecedented international cooperation for data-sharing and more, there is hope that a treatment for COVID-19 is possible. What we learn in the fight against COVID-19 will prepare us for the infectious diseases that await us. 

The authors would like to thank Bret Greenstein, SVP of Cognizant’s AI & Analytics Practice, for his contributions to this post. Additional contributors include Prasad Subramani and Kapila Monga, Cognizant AI & Analytics Client Partners serving the life sciences and healthcare industries, respectively.

This blog is part of our special report on the future of infectious disease. Stay tuned for more blogs on this topic.

Visit our COVID-19 resources page for additional insights and updates.

Digital Business & Technology, Special Report The Future of Infectious Disease AI, artificial intelligence, coronavirus, COVID-19, evolutionary AI, infectious disease, machine intelligence, Machine Learning, pandemic

Babak Hodjat

Babak Hodjat is Vice President of Evolutionary AI at Cognizant, and former co-founder and CEO of Sentient. He is responsible...

Read more

Brian Williams

Brian is Cognizant’s Chief Digital Officer for Life Sciences and is responsible for designing digitally enabled solutions to...

Read more

Related posts

Stay relevant

The Modern Business newsletter delivers monthly insights to help your business adapt, evolve, and respond—as if on intuition

Subscribe now