Bengaluru startup tests dog-and-AI system for cancer clues in breath samples

A person breathes into a specially designed mask for about 10 minutes; the sealed sample then goes to a Bengaluru laboratory, where a trained dog examines it.

Dognosis is developing a system that uses dogs’ reactions to volatile organic compounds, or VOCs, in breath samples as possible signals of cancer. Sensors record how each dog responds on a specially designed platform, and AI and machine-learning models analyse that data. The dogs do not meet the people whose samples they assess.

The company says its working sessions for the dogs are limited to about 30 to 45 minutes a day. Its Phase-2 research over the past two years involved more than 1,500 participants and reported accuracy of about 90% across seven major cancer types. Oncologist Dr Swaratika Majumdar, associated with the project, said trials showed roughly 90% sensitivity and specificity.

The study has involved six hospitals, including Karnataka Medical College and Research Institute, Karnataka Cancer Therapy and Research Institute and Radon Oncology in Hubballi. Researchers collected about 1,000 initial samples and plan larger population screening. Cancer specialists have said much bigger studies are needed to establish whether the method can reliably screen patients in real-world settings.