AI can aid faster, cost-effective diagnosis but clinical validation and data safeguards needed: experts

Artificial intelligence (AI) can improve disease detection, expand access to healthcare, and reduce diagnostic costs, but its adoption must be backed by clinical validation, data protection and human oversight, experts said in a panel discussion in Chennai on Tuesday (October 13, 2026).

The discussion on ‘Artificial Intelligence-Driven Healthcare: Innovation, Ethics, Accessibility and Sustainability’ was held as part of The Hindu’s Sustainability Dialogues 2026, presented by Saveetha Institute of Medical and Technical Sciences (SIMATS). The panellists were Sindhuja Giritharan, Clinical Lead, Digital Health and Patient Engagement, Sundaram Medical Foundation; Richa Pant, Director, Research, Deep Tek; Reuben Fernandes, Founder & CEO, Skyfire Applied Intelligence Pvt. Limited (Atom360); and A. Hariharan, Associate Professor, Department of Transfusion Medicine, Saveetha Medical College, SIMATS.

Ms. Pant said medical imaging was among the areas where AI was making a significant difference, as digital images offered standardised data from which systems could learn patterns.

“What I have observed through my research is that AI is helping reach larger populations, particularly in areas where there is a severe shortage of radiologists and other specialists. For instance, in a study we conducted in six districts of Tamil Nadu, we saw first-hand how implementing AI in TB screening could help reach more people,” she said.

AI could also help hospitals prioritise critical cases rather than process scans strictly in the order in which they arrived, enabling patients requiring urgent attention to receive treatment faster, she added.

Mr. Fernandes highlighted the potential of AI to address the low uptake of cancer screening. He said less than 1% of the population was screened for cancer, while many patients sought medical attention only after symptoms appeared, by which time the disease could have advanced to Stage III or IV.

He said falling computational costs and wider smartphone penetration could make AI-assisted screening more accessible. Smartphone-based tools could help identify suspicious oral lesions and direct patients to specialists for further assessment, Mr. Fernandes said, adding that deploying AI models directly on phones could also reduce dependence on internet connectivity in areas with limited infrastructure.

Dr. Giritharan said AI should enable doctors to devote more time to patients rather than replace the human aspects of healthcare.

AI could assist with documentation, summarising medical histories, tracking medication adherence and processing clinical information. However, doctors would still need to explain diagnoses, understand patients’ individual circumstances and help them make informed treatment decisions, she said.

“A human has to tell a human,” Dr. Giritharan said, stressing the importance of personal communication, particularly when patients were anxious about a diagnosis or struggled to follow medical advice because of financial or family circumstances.

She also called for greater emphasis on preventive healthcare. AI-assisted reminders and digital follow-ups could help doctors monitor patients with chronic conditions such as diabetes between consultations, identify difficulties in adhering to treatment and intervene when necessary, Dr. Giritharan added.

Mr. Hariharan, said there was a need for clinicians to understand the limitations of AI systems and recognise when their recommendations should not be followed. AI literacy among medical professionals was equally important, he added. He also cautioned that patient information could not be shared indiscriminately with AI systems. While India’s Digital Personal Data Protection Act and medical device regulations provided relevant frameworks, he said clearer safeguards for AI use in healthcare were needed.

On patient privacy, Mr. Fernandes said AI tool developers should avoid accessing identifiable information, such as names, phone numbers, and locations, wherever possible. Unique identifiers could allow companies to process medical data without directly accessing patients’ identities, he said.

Raising another ethical concern about whether historical patient records collected under existing consent arrangements could be used to train AI models, Dr. Giritharan said healthcare institutions needed to examine whether the consent originally obtained adequately covered such use.

The session was moderated by Zubeda Hamid, Senior Assistant Editor, The Hindu.