Imagine it’s late at night, and a woman feeling a sudden, unusual pain rushes to the emergency room, fearing the worst about a possible breast cancer recurrence. The hospital is bustling even at this hour, with patients awaiting attention. This situation paints the perfect reality of India’s healthcare infrastructure. Understaffed and slow. Just not enough personnel to meet the number of patients.
Artificial Intelligence (AI) emerges as a promising saviour at this crucial intersection of urgency and inefficiency. An alternative that every sector and industry – from healthcare, education, and media is ready to deploy – as a solution to India’s problematic allocation of budgets. In this case, AI is being made ready to be crowned as the messiah of a sector where there’s one doctor for every 834 patients1 and just 0.5 public hospital beds for every 1000 people in India2.
Public perception & the AI lifecycle
To more deeply understand this union of AI and healthcare, let’s first understand the AI lifecycle that may be initiated when you visit a hospital and provide your medical data for check-ups.
Let’s continue with my hypothetical breast cancer example to understand this.
Breast cancer, a prevalent concern for many women, typically involves rigorous screening processes like mammography, which can be stressful and often, due to human error or limitations, not entirely foolproof. AI steps into this high-stakes arena with the promise of precision. By employing advanced algorithms that analyse mammograms, AI tools help detect early signs of breast cancer more accurately and swiftly than traditional methods. This expedites the diagnostic process and potentially saves lives by catching cancer in its nascent stages.
This entire cycle — from the testing to the results — is what we can constitute the AI lifecycle.
- Data Sourcing: In the emergency room scenario, the digital image serves as critical data for AI analysis as soon as a mammogram is taken. This is where the AI lifecycle begins, with the collection of essential medical images.
- Data Cleaning and Labeling: Next, these images are processed to ensure clarity and accuracy. AI systems require high-quality, well-annotated images to learn effectively; hence, any artefacts or irrelevant information is removed, and essential features are labelled.
- Model Training and Testing: AI models are trained to detect and differentiate between benign and malignant tumours using thousands of such labelled images. They undergo rigorous testing to validate their accuracy, ensuring they provide reliable support to radiologists.
- Model Deployment: Once proven effective, these AI models are deployed in hospital settings. They assist radiologists by providing a second, highly detailed review of the mammograms, helping to identify potential breast cancers with greater precision.
- Ongoing Monitoring and Updating: Like medical knowledge, AI models are regularly updated with new data and research, ensuring they remain effective and current.
Sounds like a story out of a fairytale.
The ‘potential’ of AI
AI in healthcare represents hope, potentially accelerating diagnostics and streamlining patient management processes to compensate for infrastructural and human resource deficits. The integration of AI can significantly reduce waiting times, automate routine tasks, and assist in complex medical decisions. The urgency for such technologies is underpinned by the substantial investment growth in AI across India, which saw a leap of over 109 per cent in 2018 to $665 million3, with projections reaching up to $11.78 billion by 2025.
However, AI is not a monolith but has diverse applications in healthcare. The spectrum of AI applications is vast, from AI-powered chatbots providing mental health support to sophisticated algorithms capable of diagnosing diseases from imaging data. Each type of AI comes with its own training data and methodological frameworks, creating a layered and intricate landscape.
And so, despite AI’s potential, its deployment is not without challenges. The first step, i.e. data sourcing, involves collecting and preparing data to train AI models. And it’s a process fraught with ethical and privacy concerns.
The tech x healthcare integration is nascent in India, and the AI boom has only expedited the merger. While AI has always been used for research purposes, deploying it for diagnoses and patient care is new. And so, while the research section is decently regulated, the data sourcing part for patient care is not.
This lack of stringent regulations and the ‘data brokers’ market make it seem like Wild Wild West. According to a Mint report, “Healthcare data can fetch up to $250 per record in the black market, compared with the next highest valued record of $5.40 for credit or debit cards.”4
In a country that’s only just beginning to gain digital equality and penetration, lack of data literacy is a huge problem even otherwise. The sheer amount of exposure and data sharing that invisibly happens daily on the internet raises the stakes for data privacy even more.
Consent, amidst all this, emerges as one of the biggest concerns in this fast-paced AI race.
Challenges in data utilisation and dependency
A recent poll by Youth Ki Awaaz, taken by 300 respondents, revealed that 60% of participants were reluctant to share their personal medical data, which is essential for training effective AI systems. This solid public sentiment stresses the urgent need for robust ethical guidelines and transparent data handling practices to build public trust and ensure the successful integration of AI in healthcare systems.
Notable incidents like the cyberattack on AIIMS, Delhi, highlight the vulnerabilities associated with data management. India has been identified as the second-most affected country by data broker breaches, according to a report by data removal service Incogni5, emphasising the widespread impact of these breaches on personal security.
The reliance on user consent for data usage poses another hurdle. Current guidelines, such as those from AISEMR, focus on biomedical research and do not extend to healthcare practices, creating a regulation gap that directly impacts AI training in medical settings. Moreover, although well-intentioned, Ayushman Bharat Digital Mission’s privacy policies often clash with the practical needs of AI deployment, leaving private hospitals to depend heavily on individual consent.
To sum it up, the use of data in the healthcare context is weakly overseen, with archaic rules unfit in times where AI-based tools are used.
Relying on Western data to train the systems is often posed as a solution. Still, given the cultural and epidemiological differences, the risk of importing biases and inaccuracies in medical diagnostics and treatment plans is just too high.
As we envision the future of AI in healthcare, the path forward must involve crafting localised AI strategies that respect privacy, enhance data security, and are tailored to the specific health needs of the Indian population.
And it might be worthwhile to question where our priorities lie: ensuring we have a tech health ecosystem or fixing the problem more ground up and allocating more budgets for it? In another recent poll conducted by Youth Ki Awaaz, 62% of respondents believe the government should prioritize affordable medicine, while 35% think the focus should be on more hospitals. Notably, only 2.63% see online health services—a proxy for tech-driven solutions—as a priority.
India has never fully and honestly committed to making the roots of healthcare systems stronger, and it’s time we do that, instead of finding a cop out in AI. As for AI, with its current challenges unaddressed, the possibility of a fully integrated AI-powered healthcare system in India remains just that—a dream.
References:
- Patel, Shivam. India Builds More Hospitals as Population Surges but Doctors in Short Supply | Reuters, www.reuters.com/world/india/india-builds-more-hospitals-population-surges-doctors-short-supply-2023-05-10/. Accessed 23 July 2024.
↩︎ - Srinivasan, Aravindan. “Health Infrastructure, Capacity Building & Investments: The Focal Points to Improve Healthcare in India – et Healthworld.” ETHealthworld.Com, 20 July 2023, health.economictimes.indiatimes.com/news/industry/health-infrastructure-capacity-building-investments-the-focal-points-to-improve-healthcare-in-india/101970232. Accessed 23 July 2024.
↩︎ - “Ai for Societal Transformation: India’s Vision In Action.” AI for Societal Transformation: India’s Vision in Action, www.psa.gov.in/mission/artificial-intelligence/34. Accessed 23 July 2024.
↩︎ - Choudhury, Abhijit Ahaskar &Moumita Deb. “Indian Hospitals, Clinics, Labs Selling Data without Consent.” Mint, 18 Jan. 2022, www.livemint.com/companies/indian-hospitals-clinics-labs-selling-data-without-consent-11642532931402.html. Accessed 23 July 2024.
↩︎ - Shetty, Soham. “India among Top 5 Countries Most Affected by Data Broker Breaches, Report Reveals – CNBC TV18.” CNBCTV18, 10 Mar. 2023, www.cnbctv18.com/news/india-among-top-5-countries-most-affected-by-data-broker-breaches-report-reveals-incogni-16139021.htm. Accessed 23 July 2024.
↩︎