BENGALURU: The test begins with something almost deceptively simple: A person breathes into a specially designed mask for about 10 minutes.There is no needle, no scan and no elaborate procedure for the person being tested. What the mask captures, however, is something scientists believe may carry clues about disease — volatile organic compounds, or VOCs, present in human breath.The mask is then sealed and taken to a laboratory. This is where the unusual part begins.Instead of immediately going under a machine, the sample is presented to a specially trained dog.At a Bengaluru-based startup, dogs are being trained to distinguish the chemical signatures associated with cancer. Their reactions are then recorded through sensors and fed into artificial intelligence systems, turning what would otherwise be a dog’s instinctive response into data that researchers can analyse.The idea sounds almost like something out of a science-fiction laboratory: A dog sniffs a breath sample while an AI system watches.But behind it is a simple proposition — can the extraordinary sensitivity of a dog’s nose be combined with technology to create a faster, non-invasive way of screening for cancer?Dognosis, the Bengaluru startup working on the technology, believes it can.
The samples are collected through masks and taken to the company’s laboratory, where trained dogs examine them under controlled conditions. (AFP image)
A breath sample, a dog and thousands of sniffs
Dogs have long been known for their exceptional sense of smell. The Bengaluru project is attempting to put that ability to work in a controlled medical setting.According to Dognosis CEO Akash Kulgod, breath samples contain VOCs that can carry signals indicating whether a person has a disease.The samples are collected through masks and taken to the company’s laboratory, where trained dogs examine them under controlled conditions.The dogs do not meet the people whose samples they are examining.Instead, each sample is presented in the same way, allowing researchers to observe how the animals respond. Sensors record their behaviour, while AI and machine-learning models analyse the resulting data.Suba, who heads research and development at Dognosis, said uniformity is critical to the process.Every sample, she explained, is presented in exactly the same manner so that the dog’s response can be recorded precisely. The resulting data is then processed through AI algorithms, with the aim of converting an inherently subjective behaviour — how a dog reacts to a smell — into objective measurements.That is the key difference between simply saying a dog can “smell cancer” and attempting to build a medical technology around the ability.
The dogs are not the machine. They are part of it
At the centre of the experiment are the dogs themselves.Dognosis has brought in trainers, including international canine specialists, to prepare the animals for the work. In the company’s testing environment, the dogs are trained to associate a particular scent with a reward.The animals may respond differently — some run towards the feeder, some freeze, while others scratch at the sample in excitement. For the researchers, the exact style of response matters less than being able to consistently record it.In the company’s setup, the dogs work on specially designed platforms equipped with automatic feeders. Their reactions are captured through sensors and converted into data.The startup says its dogs can analyse a large number of samples, with Kulgod describing the animals as capable of thousands of sniffs in an hour.The dogs, however, are not being treated simply as biological instruments. The company says their working time is limited to around 30 to 45 minutes a day.
From Hubballi hospitals to Bengaluru labs
The research has extended beyond the startup’s laboratory.Three hospitals in Hubballi — Karnataka Medical College and Research Institute, Karnataka Cancer Therapy and Research Institute, and Radon Oncology — have collaborated on the breath-based cancer detection study.Sample collection began at KCTRI in April 2024. Patients with cancer, people undergoing other medical treatment and healthy participants were approached for the research.The process remains deliberately simple: participants are informed about the study and asked to breathe into the specially designed mask for around 10 minutes.Researchers said about 1,000 samples were initially collected as proof of concept. The project then sought thousands more samples to generate clinical data and train the dogs, with a much larger population screening planned later.The Bengaluru-based team has also worked with hospitals in Bengaluru, taking the total research effort across six hospitals.The company says its Phase-2 work over the past two years involved more than 1,500 participants and reported accuracy of around 90% across seven major types of cancer.Dr Swaratika Majumdar, an oncologist associated with the project, said the technology had shown around 90% sensitivity and specificity in the trials.Those numbers are promising, but researchers outside the project have urged caution.Cancer specialists have described the findings as an encouraging early signal while stressing that much larger studies are needed before the technology’s usefulness for patients can be established conclusively.That distinction matters. Detecting a pattern in a controlled study is one thing; proving that the same system can reliably screen large populations in the real world is another.
The idea is older than AI
The Bengaluru experiment may sound futuristic, but the underlying fascination with dogs and disease is not new.Scientists have long investigated whether dogs can detect human illnesses through smell. Research has explored their ability to identify chemical changes associated with various diseases.Dogs have also attracted interest as models for understanding human disease itself.Research into the canine epigenome has found similarities between molecular changes associated with cancer in dogs and those observed in human cancers. Dogs can develop cancer spontaneously, as humans do, making them useful subjects for studying disease mechanisms.The connection between humans and dogs, in other words, is not merely about companionship. Their biology and extraordinary sense of smell have increasingly become subjects of scientific investigation.What is different in Bengaluru is the attempt to add another layer to that relationship: artificial intelligence.
From a dog’s reaction to a number
A dog’s nose can detect something humans cannot easily describe.But a medical test needs more than an impressive biological ability. It needs consistency, measurable results and a way to interpret those results.That is where AI enters the picture.Instead of relying on a trainer simply observing whether a dog reacts to a sample, the Bengaluru system attempts to capture the animal’s behaviour through sensors and analyse it computationally.The goal is to turn a biological response into a repeatable dataset.It is an unusual partnership: the dog provides the biological sensitivity, the sensors record what happens, and AI attempts to find patterns in the resulting data.If successful, the technology could potentially allow the dogs’ ability to be used in a more standardised way than traditional scent detection.
A simple test for a growing problem?
India’s projected cancer numbers give the experiment a much larger context.The ICS has warned that the country could be facing nearly 2.5 million new cancer diagnoses annually by 2045, up from around 1.5 million currently.For doctors and public-health experts, the challenge is not only what happens after a diagnosis. It is how many people can be identified early enough to benefit from treatment.Dr Urvashi Prasad, cancer survivor and former director at NITI Aayog, has pointed to regional disparities, gaps in cancer data and financial barriers as continuing obstacles to timely diagnosis and treatment.Dr Monika Puri, public health consultant and former WHO official, has argued that prevention, screening and continuity of care need to be integrated into primary healthcare systems if cancer care is to become more equitable.A breath-based test, if eventually proven reliable, could potentially fit into that larger idea of accessible screening.But that remains a possibility, not an established reality.
From laboratory experiment to cancer screening
For now, the technology remains a work in progress.Dognosis says it hopes to move towards commercial use, but the larger challenge lies in establishing the test through broader clinical validation and regulatory approval.The promise, however, is difficult to ignore.Imagine a cancer screening system in which the first step is not a scan or an invasive procedure but simply breathing into a mask. The sample goes to a laboratory, where a trained dog detects chemical signals invisible to the human nose, while sensors and AI translate the dog’s reaction into measurable data.It would not mean replacing doctors, biopsies or imaging.Instead, the ambition is to create another layer — one that could identify people who may need further testing before cancer has progressed.For Bengaluru’s biotech ecosystem, that is perhaps the most intriguing part of the experiment.And at the centre of it all is a dog, quietly sniffing for a signal that humans cannot smell.
