This Bengaluru startup believes man’s oldest friend can sniff out cancer 

A Bengaluru startup is betting that one of medicine’s oldest diagnostic tools may have been sitting under our noses all along. 

Dognosis is beginning a 10,000-person trial in India to test whether trained dogs, combined with artificial intelligence, can identify chemical signatures associated with cancer from a person’s breath. The ambition is straightforward: build a cheap, non-invasive screening layer that can flag people who may need conventional diagnostic tests before symptoms appear. 

Can canines can cancer? 

The idea sounds almost whimsical. The science is considerably more serious. Dogs possess an extraordinary ability to detect volatile organic compounds, or VOCs, produced by metabolic processes in the human body. Some of those compounds appear to change when cancer develops. Dognosis trains dogs to identify these scent patterns in breath samples, while its statistical and AI systems help turn the dogs’ individual responses into a clinical signal. 

The company’s recently published Phase II study, in the Journal of Clinical Oncology, provides a stronger foundation for the experiment than the early headlines might suggest. The study enrolled 3,275 people across six hospitals in Karnataka, with 1,502 participants in the testing cohort. The system recorded 90.8% sensitivity and 91.3% specificity across seven major cancer groups, while sensitivity for stage I and II cancers was 90.6%.  

The new 10,000-person exercise is therefore important for a different reason. A promising case-control result is not the same thing as a population-screening test. Dognosis now has to demonstrate that its approach can work reliably at considerably greater scale, particularly among people who may not have cancer at all. 

That challenge is being pursued elsewhere, too. 

There are global precedents 

In the US, SpotitEarly is developing a remarkably similar model, combining trained dogs with AI and breath analysis. The company says it has raised more than $20 million, including $16.7 million in equity financing and $3.6 million in grants. It is conducting major US validation studies and is targeting breast, lung, colorectal and prostate cancers.  

Dealroom currently estimates SpotitEarly’s enterprise value at roughly $81 million to $122 million. That is an estimate rather than a disclosed transaction valuation, but it offers a useful indication of how investors are beginning to price the category.  

There is also a broader breath-diagnostics industry developing around the same scientific premise, even without dogs. UK-based Owlstone Medical, for example, is developing its Breath Biopsy platform to identify disease through VOCs. It has raised more than $150 million since its founding, including a $58 million Series D and a $27 million first close of its Series E in 2025.  

Dognosis itself remains much earlier in the investment cycle. Publicly available funding information puts its disclosed capital at around $2 million, with backing including Boost VC, 1517 Fund, MBX Capital, House Fund and Accel India. There is no reliable publicly disclosed valuation for the company, so attaching a number to it would be speculative.  

That disparity is revealing. The global opportunity is attracting tens of millions of dollars, while Dognosis is attempting to establish whether the model can work economically and clinically in a country where conventional cancer screening remains uneven. 

That could ultimately be its biggest advantage. A laboratory full of highly trained dogs is hardly the obvious image of scalable healthcare. But if breath samples can be collected at home, transported centrally and assessed using a repeatable canine-AI workflow, the economics could look very different from scans, biopsies and sophisticated laboratory tests. Dognosis says it is targeting affordable home testing in India and eventually the US. The company reportedly hopes to establish an American facility around 2027-28.  

The dogs, in other words, may not be the product. They may simply be the biological sensor. The real business could be the technology that teaches machines how to understand what those noses already know.