BY THE OPTIMIST DAILY EDITORIAL TEAM
A company in India is reporting more than 90 percent accuracy in detecting seven types of cancer from breath samples. The diagnostic engine is artificial intelligence, trained on data from dogs.
The 15 dogs at Dognosis, a startup based near Bengaluru, are not making the call. They wear helmets and harnesses fitted with sensors that track their breathing, brain activity, and movement as they sniff sealed samples. The AI processes that biometric data and looks for patterns in how the animals respond to cancer-related odor compounds. The dogs are the instrument; the algorithm reads it.
A 10-minute breath test, no blood draw required
Patients breathe into a cotton mask for about 10 minutes. The sealed sample goes to the lab, and the patient never meets the dogs.
Dognosis says a study of more than 1,500 participants across six hospitals found accuracy above 90 percent across seven cancer types. That figure comes from the company and has not been independently verified. One company-reported study does not make this a reliable diagnostic method, and researchers are saying as much.
Oncologist Santhosh Kumar Devadas put it plainly: “The study is promising, as an early signal. But they need to do the study with a much larger group before we make any definitive conclusions.”
Why this matters in India
A large share of cancer diagnoses in India come only after the disease has already advanced, partly because accessible, non-invasive early-screening options are thin on the ground. A breath test that flags people who may need further testing, without requiring imaging, blood draws, or specialist access at the first step, would fill a real hole.
Dognosis is pitching the tool as a first-pass filter, not a replacement for conventional diagnostics, and says it is pursuing further trials toward potential commercial approval.
How the AI is trained
The AI is not learning to detect cancer directly. It is learning what a dog’s body does when it encounters specific odor compounds associated with cancer in human breath; those involuntary responses, recorded through sensors and converted into data, are what the system trains on. The goal is to take something dogs already do and turn it into something consistent and repeatable at scale.
Researchers are clear that larger, independent studies will be needed before this holds up for clinical use. The 90 percent figure is striking, and it also comes from a single study run by the company building the product. Both of those things are true.
Did this solution stand out? Share it with a friend or support our mission by becoming an Emissary.
