AI Surpasses Human Performance in Real-Time Medical Diagnosis, Study Confirms

NEW DELHI — In what experts are calling a watershed moment for artificial intelligence in medicine, a new study published in the New England Journal of Medicine on Monday revealed that DeepDiagnose, an AI model developed by Bengaluru-based startup MedCore AI, correctly identified rare conditions in 94.7% of cases — compared to 81.2% for a panel of senior physicians across three major Indian hospitals.

The research analyzed over 48,000 patient cases at AIIMS New Delhi, Tata Memorial Hospital Mumbai, and Fortis Gurugram — the largest clinical AI study ever conducted in South Asia.

“We did not expect the gap to be this wide,” said Dr. Arun Mehta, lead author and professor of diagnostic medicine at IIT Bombay. “The model showed particular strength in identifying overlapping symptom profiles that often confound even experienced clinicians.”

Reactions from the Medical Community

The study sparked immediate reactions worldwide. India's Ministry of Health released a statement saying it would form a working group to evaluate regulatory frameworks for AI diagnostic tools in public hospitals by September 2026.

Not all experts are enthusiastic. Dr. Fatima Malik of the Indian Medical Association cautioned that clinical judgment involves far more than pattern recognition. “A diagnosis is a conversation, not a calculation,” she wrote in a rebuttal on the IMA blog.

Instagram Post