Multilingual AI for Diagnostics: A Leap Towards Accessible, High-Quality Healthcare in India
By Arunima Rajan
In an interview with Arunima Rajan, Prof. Himanshu Sinha, faculty, Department of Biotechnology, IIT Madras and Wadhwani School of Data Science and AI, IIT Madras, explains how AI and multilingual LLMs are transforming diagnostics to bring accessible, high-quality healthcare to every corner of India.
With so much innovation on the horizon, which emerging research areas do you believe might become true game-changers in enhancing diagnostic accuracy and accessibility?
India's healthcare sector stands at the cusp of a transformative era, with Large Language Models (LLMs) offering promising solutions to enhance accessibility and quality across the country’s diverse linguistic and cultural landscape. These AI-powered systems have the potential to revolutionise healthcare delivery, particularly in underserved and rural areas, by providing medical information and support in multiple Indian languages. This multilingual capability is crucial in a country with 22 official languages and hundreds of dialects, as it enables telemedicine platforms to overcome language barriers and facilitates remote consultations in local languages.
In rural India, where healthcare infrastructure is often limited, LLMs can bridge critical gaps by deploying AI-powered diagnostic tools in rural clinics. These tools can assist healthcare workers in making accurate diagnoses without on-site specialists, a game-changer for areas with limited medical expertise. Voice-based AI models and IVR systems can expand the use of regional languages in healthcare communication, ensuring accessibility even in areas with limited internet connectivity. This is particularly beneficial for rural populations who may have lower literacy rates or limited access to smartphones.
Prof. Himanshu Sinha, faculty, Department of Biotechnology, IIT Madras
The potential of LLMs extends to personalised care and decision support, analysing patient data to create tailored treatment plans that consider individual medical history and genetic information. This capability is especially valuable in a country with diverse genetic backgrounds and varying health challenges across regions.
However, implementing LLM-based solutions in Indian healthcare is not without challenges. Data privacy and security concerns are paramount, especially when handling sensitive health information in a country that is still developing its data protection framework. There's also a pressing need for high-quality, diverse datasets in multiple Indian languages to train effective AI models. Ensuring cultural sensitivity and proper localisation is crucial to making AI-generated content appropriate for different regions and communities across India.
Looking ahead, the full potential of LLMs in Indian healthcare can be realised by developing specialised models trained on Indian medical guidelines and practices. Implementing continuous feedback mechanisms involving community health workers and healthcare professionals will be key to improving AI model accuracy and relevance.
Exploring innovative deployment models, such as offline-capable voice AI, can help overcome infrastructure limitations in remote areas.
Bridging the gap between groundbreaking research and practical diagnostic tools is a challenge in itself. What are the key hurdles in this process, and how might academic and industry collaborations help overcome them?
Collaborations between IIT Madras and industry partners are playing a crucial role in overcoming these challenges and delivering these technologies as digital public goods.
IIT Madras has established the Centre for Responsible AI (CeRAI), which focuses on developing ethical and accountable AI-driven solutions. CeRAI collaborates with industry partners like Roche Diagnostics to advance research in AI for healthcare, ensuring that the developed technologies are innovative and ethically sound. One of the areas of focus of this partnership is in-vitro diagnostics.
The AI4Bharat project, a collaboration between IIT Madras and the ekStep foundation, is pioneering the development of multilingual LLMs tailored for Indian languages. This initiative has released models like IndicBERT and IndicBART, which are specifically designed to handle the complexities of Indic languages. AI4Bharat and other initiatives at IIT Madras are committed to releasing their developments as open-source contributions. This approach democratises access to these technologies, allowing wider adoption and further development by the community.
Could you share a glimpse into any current projects or breakthroughs from your research that you feel could significantly alter the way diagnostics are approached in India?
IIT Madras has been at the forefront of developing cutting-edge healthcare technologies that are poised to significantly alter the approach to diagnostics and healthcare delivery in India, with a particular focus on maternal and child health.
Here's an overview of our most notable projects and breakthroughs:
AI and Responsible Healthcare
As already noted, CeRAI at IIT Madras is collaborating with industry leaders like Roche Diagnostics to develop ethical and accountable AI-driven healthcare solutions, including in-vitro diagnostics and ethical frameworks for transparent decision making.
We have also made significant strides in improving maternal and child health outcomes. This includes the 'Garbhini-GA2' AI model, which accurately determines fetal age in the second and third trimesters of pregnancy:
Reduces gestational age estimation errors by almost three times compared to Western models
Improves care delivered by obstetricians and neonatologists
Helps determine precise delivery dates
The Dharani Dataset, from Sudha Gopalakrishnan Brain Centre, headed by Prof. Mohanasankar Sivaprakasam, unveiled in December 2024, provides the world's first publicly accessible 3D high-resolution digital images of the human foetal brain:
Facilitates early detection of brain abnormalities during pregnancy
Supports research into conditions like autism spectrum disorder and epilepsy
Enables the development of AI-based diagnostics for prenatal care
By combining cutting-edge research with practical applications, IIT Madras is paving the way for a new era of healthcare in the country, with a strong focus on improving maternal and child health outcomes.
Got a story that Healthcare Executive should dig into? Shoot it over to arunima.rajan@hosmac.com—no PR fluff, just solid leads.