×

How India is using Artificial Intelligence to combat COVID-19

Decision makers using computer simulations to understand how the pandemic evolves

Artificial Intelligence (AI) has been one of the biggest technology success stories of the past decade. As the COVID-19 pandemic spread across the globe, researchers and entrepreneurs stepped up to devise new ways to combat it. From understanding and preventing the spread of the virus to diagnosing and treating it, startups and established technology companies in India are actively leveraging AI to support this fight.

Modeling the pandemic

Decision makers have increasingly relied on computer simulations to understand how the pandemic situation will evolve over time. TCS, in collaboration with Pune-based Prayas Health Group, is using digital twins to forecast the spread of COVID-19 in urban districts. A digital twin is a virtual computerised model of a physical system that takes real-world data as input and predicts the future evolution of the system.

"Macro models don't work well in countries like India which have high heterogeneity. So we developed ward-level digital twins that modelled the spread of the disease as a function of the number of proximal contacts, average duration of contacts, people and place characteristics, and population demographics like age, gender, comorbidities, etc,” says Vinay Kulkarni, distinguished chief scientist, TCS. “The model predictions closely match the observations reported by city corporations and empower the administration to take better locality-specific decisions."

Prevention

Ensuring people wear face masks and follow social distancing is expected to be a major challenge for organisations. AI is being used to monitor live CCTV feeds and instantly report violations of guidelines to safety administrators.

“It is critical that a safe ecosystem is created for businesses and schools to re-open as soon as possible in spite of COVID-19,” says Atul Arya, CEO of Blackstraw, an AI company. “Our AI-powered health risk management system developed jointly by Blackstraw and Bharat Forge not only enables safety of humans and compliance with government guidelines, but also drives long-term behavioural changes that are crucial to live by in the new normal.”

Diagnosis

The reverse transcription polymerase chain reaction (RT-PCR) test is considered the gold standard for COVID-19 diagnosis. Due to long processing times, many hospitals use chest X-ray and CT scans for screening patients. Radiology scans are also useful in monitoring progression of the disease and assessing the degree of lung infection. AI has made rapid advances in the last few years in diagnosing tuberculosis amongst other pathologies from radiology images. Enterprising medical startups were quick to repurpose their TB solutions for COVID-19.

“We do not believe AI can yet replace radiologists. Our AI solutions are, instead, designed to augment and assist them,” says Dr Amit Kharat, a radiologist and co-founder at DeepTek. “Our AI models have been used in the field to report over 80,000 X-rays for a government-run TB screening program. We are using the same base technology for COVID-19 screening.”

By decreasing radiologist efforts and improving reporting times, such AI-enabled solutions help in making timely and accurate diagnosis affordable for everyone.

Treatment

TCS is using AI simulations to synthesise molecules and discover new drugs to fight the virus. From a candidate set of 50,000 molecules, their simulations selected 31 molecules that are now undergoing trials as potential cures.

A new untested drug needs to pass extensive clinical trials on human patients before approval. Hence, the short-term focus of pharmaceutical companies has been on repurposing existing drugs that have already cleared trials for COVID-19 treatment. Innoplexus is using AI to crunch through a massive dataset of available drugs to identify safe drugs that could disrupt the functioning of the virus.

Regulations and policy

Hospitals and governments have tempered their enthusiasm for AI with caution. AI systems are brittle and work only on the specific datasets that were used to build them. They need to be extensively tested on real-world data before being adopted in practice.

“It is difficult to trust AI since it cannot explain its predictions,” says Sahil Deo, an AI-policy expert and co-founder of CPC Analytics. “There is also the question of who is to be held accountable if the AI mispredicts. We need a strong regulatory framework before we can widely adopt AI in decision making.”

(The author has a master’s degree in computer science from University of California, Berkeley and is currently pursuing a Ph.D. in Quantum Artificial Intelligence)