According to a global study by PwC India, the highest increase in the use of AI during COVID-19 times has been witnessed in India. The study indicated AI adoption in India was 73 percent in healthcare and pharma companies. Many experts cite that health emergency caused by COVID-19 has put AI on centre stage.
Addressing COVID-19 with AI
The healthcare thought leaders informed that COVID-19 response has led to challenges, opportunities and innovation at every level. AI solutions have sprung up in step with the needs of patients, physicians, hospitals, researchers, drug companies and policymakers.
Giving examples of how the AI is used to manage the pandemic response globally, Tavpritesh Sethi, Associate Professor, Computational Biology, IIIT-Delhi, informed that Blue Dot — an early alert system integrated information on infection spread with flights to create an early warning system used previously for Zika and H1N1. Another solution that emerged during this time was risk calculators, which enabled doctors and patients who want to know the risk for a fatal deterioration at home or in hospital settings as. In the diagnostic space, over 400 solutions for diagnosing COVID-19 from images have propped up during this time, some of which have been successfully deployed in the Indian settings. Automation of the supply chain is one of the lowest hanging fruits for the use of AI during the pandemic.
“AI plays a role in reducing the burden on overstretched and overworked healthcare workers and it reduces human errors due to overwork. It acts as a physicians’ assistant. In India especially, where we have an enormous lack of resources, AI should be seen as a necessity. It can also help with efficiencies, reducing costs and simplifying routine tasks,” said Devang Lakhia, Country Head – India & South Asia Hamilton Medical India Pvt Ltd.
Further elucidating on how AI is helping in managing COVID, Dr.Satish Prasad Rath, VP, Chief Innovation & Research Officer, Aster DM Healthcare, said, “COVID has expedited technology adoption in healthcare. Also, it has helped in categorising segments of AI, which were mature and ready for adoption to evolving form of AI. Use of AI to identify the key molecules at sub cellular level and ability to simulate in silico experiments on new drug target has been the most impactful effects of AI usage. Apart from vaccines, AI also has been very helpful in genomic surveillance, which has been key for successful containment for the emerging variants.”
He also added that apart from genomics, AI was found very helpful in rapid and accurate diagnostics. The globe is struggling with a low provider-to-patient ratio, which has been exaggerated with the pandemic. AI can be a savior here to complement and augment known healthcare practices. AI-powered digital surveillance has enabled proper containment zone demarcation, assessment, and appropriate resource allocation. As we move from lockdown to mobilization, this area is going to stay relevant for effective continuity of economy and business.
Experts also cite that as they were contemplating advanced data analysis techniques and integrating those algorithms into technologies and develop end to end solutions in the healthcare segment, the pandemic made them leapfrog to drive quick resolution, given the urgency of the situation.
“The pandemic has taught us the necessity of reliable data. AI model will be as good as the underlying data, and a lot of AI practitioners and researchers have struggled with this need. This has also opened up opportunities to think about the world around us as data. Our own lab has been working on triangulating insights from the conventional and unconventional sources of data around us. Finally, the recent onslaught of the new variants in India has taught us that we can never be over-prepared for the pandemics and need to build a better data and AI approach to healthcare during the better times as opposed to a fire-fighting response. COVID-19 has been a whirlwind that took the healthcare system by surprise. A lot of AI solutions currently under validation or regulatory approvals are likely to be deployed in the coming months,” Sethi added.
Doctors and health tech experts inform that AI is no more alien to the domain, the pandemic has emphasised that it is a necessity and not a luxury.
Embracing AI – a necessity
As per the experts, anything that is expected to become integral to the delivery of healthcare is a necessity. India has about 4.8 doctors per 1000 people on average, and some areas have less than one doctor per 1000 patients. With AI, this number is expected to reach ~6.9:1,000 by 2023. Also, the pace of AI advancements will inevitably make AI accessible first to privileged sections of society. Therefore, it is vital to think of AI as necessary to make healthcare inclusive, available, and safe for everyone.
Highlighting the importance of AI, Suthirth Vaidya, CEO – Predible, said, “Given the overburdened health system today, AI is helping healthcare staff manage their patients efficiently and triage those who need immediate attention. The same applications are being witnessed across multiple forms of digital data today, starting from radiology (X-rays and CT scans), bedside monitoring devices, and even home monitoring solutions. AI-based algorithms are able to accurately recognize the severity of the condition, predict deterioration and help identify those who required additional care or support.”
Adding to it, Sean Narayanan, CEO and Board Member, Apexon, said, “AI and ML-driven technology is helping with image scan analysis and is helping reduce the workload on hospital staff. Several programs are now available for chest screening that can highlight lung abnormalities in a chest X-ray scan and provide a COVID-19 risk evaluation faster than before. Digital medical consulting platforms are also being developed with a unified view of the patient’s medical history to provide treatment remotely and reduce the chances of hospitalization. AI-driven solutions like applications let healthcare staff know when to expect patients and inform them of the travel and waiting time involved to see/meet the healthcare provider, function based on traffic patterns and scheduling information. This helps patients decide when they may leave for the facility and reduce human contact as they would not have to wait at the healthcare provider’s office.”
Although AI tools are most sought by health experts and are seen as a powerful weapon against pandemics, particularly in prediction, diagnosis, and treatment, a debate on AI ethics is still rampant.
Ethics in AI: Needs a fine balance
The privacy of patient data encoded within the technology has left many industry experts concerned about the ethical implications of AI. At the same time, experts also point out that ethical use of data cannot be ensured by the technology itself. The responsibility for this lies with human beings.
“AI application has a vulnerability to be biased towards a particular community if the initial models aren’t constructed scientifically. The onus is on the AI scientists to be cognizant of this possibility and prevent this from the start itself. This is scientifically possible and we have to resist the temptation of getting into the market with a half-baked solution,” Dr.Rath informed.
Commenting that there are many ways in which the ethical problems around AI can be addressed, Sethi said, “Privacy-preserving AI technologies can allow contact tracing to happen without the need for data to move to any server. I believe that the bigger issue is trust. That can be solved by building technology that communicates with its users transparently, in their own language and format that they can understand.”
Informing that most AI-based solutions are being deployed on cloud servers, Vaidya, said, “It is important that technology protects patient privacy and prevents misuse of the data. Most systems today are positioned to act as an assistant to existing healthcare staff and practitioners, thereby ensuring strong controls to prevent any risk from being carried over to the patient.”
Simultaneously, the experts also voiced that AI is not a silver bullet solution; it does have real-world challenges.
“AI has challenges the most significant of which are generalizability and reproducibility. Generalizability means that models developed in one setting can be applied to a different setting without much loss of accuracy. This is a standard requirement for drugs and devices before these can be marketed. However, AI models, having to rely upon data and the noise around them, are much more difficult to be generalized. For example, a recent study showed that none of the 400 plus models developed upon CT scans during January – October 2020 were clinically deployable due to underlying biases or methodological flaws,” Sethi said.