Artificial intelligence (AI) has shown significant promise in improving access to, and effectiveness of, medical imaging in screenings and diagnostics to assist healthcare professionals. It has been shown to improve the reach and effectiveness of experts even at the secondary and tertiary levels. However, creating and validating such solutions requires data that are representative of India’s diversity of people, places, and settings. Since these factors change over time, along with the factors that cause diseases (for example: the pathogens), the mechanisms should allow for data to periodically/continuously monitor deployments and adjust.
The organisers of this workshop on ‘Indian Gold Standard Medical Imaging Datasets’ believe that a system that provides purpose-appropriate access to modern, synchronised, and standardised training and validation datasets of the Indian population can bridge the efficacy gap, unlock new discoveries, and realise the potential of AI in medical imaging to improve patient/clinical outcomes. Keeping this in mind, they convened a workshop during 13–14 March 2023 to:
● develop a shared understanding of the state of research, validation frameworks, policy development, and the role of Gold Standard medical imaging datasets to advance tuberculosis and diabetic retinopathy screenings; and
● create a network of practitioners to articulate key questions, identify priorities, and facilitate meaningful collaborations.
The workshop also coincided with the launch of the ICMR-IISc Centre for Health Data, the aim of which is to make India a leader in data-driven healthcare research and innovation. The inauguration occasion was graced by Rajiv Bahl, Director General of the Indian Council of Medical Research (ICMR), Govindan Rangarajan, Director of the Indian Institute of Science (IISc), experts from research institutes, and industry leaders from the field.
The workshop saw an energetic participation and open discussions by the nation’s leading Health AI thought leaders and experts from nine research institutions and three industry collaborators:
- Indian Council of Medical Research, New Delhi
- Indian Institute of Science, Bengaluru
- National Health Authority, New Delhi
- National Institute of Research in Tuberculosis, Chennai
- National Institute of Mental Health and Neurosciences, Bengaluru
- Post Graduate Institute of Medical Education and Research, Chandigarh
- Aravind Eye Hospital, Madurai
- Sankara Nethralaya, Chennai
- Indraprastha Institute of Information & Technology, New Delhi
- ARTPARK, Bengaluru
- KHPT, Bengaluru
- InnoWave Technologies, Chennai
The programme covered the landscape of Health AI research and promising developments, discussions exploring starting points, and the next steps. The participants touched upon translating and scaling innovations and breakthroughs to the national level, their usability in different care settings, and awareness/education among patients, caregivers, and providers. The clinicians highlighted that the way in which AI systems are integrated would be crucial. To minimise the risk and increase adoption, human-centred approaches—that empower healthcare staff to understand how a decision is made and how to incorporate this knowledge into treatment – are required.
The working sessions focussed on creating purpose-appropriate and complete Gold Standard datasets for (1) tuberculosis screening and (2) diabetic retinopathy screening.
- Tuberculosis: India has 2.4 million notified tuberculosis (TB) patients. It is estimated that about 40% of the country’s population has latent TB infections. The TB care pathways are not straightforward and are more complex on the ground than they may seem. The chest x-ray and cough sound-based TB screening research areas are popular among data scientists and AI researchers.
Latent TB infections, as a topic, garnered much interest from the workshop participants. The plan is to schedule follow-up discussions with National Institute of Research in Tuberculosis (NIRT) scientists and team members to understand, define, and prioritise latent TB research areas.
- Diabetic retinopathy: The most common eye complication of diabetes is diabetic retinopathy (DR), along with cataracts and retinal vascular disorders. India is reported to have the second largest population of diabetes (78 million), and one in every ten diabetes patients is at risk of losing sight/vision due to DR. This is preventable with early DR screening initiatives/programmes. However, the scaling up of such programmes is challenging due to the lack of infrastructure, non-availability of hardware, skewed ratio of trained healthcare professionals and patients, and a missing connection or network b/w screening site and referral centres.
Nevertheless, compared to TB, DR screening is a relatively better-defined research area, and the opportunity here is to co-create purpose-appropriate, representational Gold Standard datasets that can provide solutions to increase access in remote/resource constrained areas and to improve the screening outcomes. There is ongoing work with the Post Graduate Institute of Medical Education and Research, Aravind Eye Hospital and Sankara Nethralaya to explore this opportunity further.
The importance of purpose-defined, highly-curated, and representational Gold Standard datasets in discovering new and innovative technology breakthroughs that could impact people in India and across the world was highlighted. The plan is to let this workshop evolve into an ongoing forum for interactions and catalysing action collaborations.