![\"\"](\"https:\/\/yogaesoteric.net\/wp-content\/uploads\/2026\/09\/dog.jpg\")
<\/p>\nA simple breath test, specially trained dogs, and technology: what initially sounds like an unusual research project could develop into a new method for early cancer detection. The Indian-American company Dognosis is working on combining the extraordinary sense of smell of dogs with sensors, brainwave measurements, and machine learning.<\/p>\n
<\/p>\n
The idea: the dog’s visible behaviour alone should not determine whether a sample is classified as potentially cancerous. A specially adapted helmet and harness additionally record brain activity, breathing, and body language. Algorithms then learn to interpret these signals to determine whether the dog has detected a specific disease odour.<\/p>\n
Dognosis co-founder Akash Kulgod describes the equipment as a kind of \u201cIron Man suit for dogs<\/em>.\u201d The project has since been backed by a Phase II trial with 1,502 participants, published in the Journal of Clinical Oncology in 2026.<\/em><\/p>\n Diseases leave a smell<\/strong><\/p>\n The ability of dogs to smell diseases has been the subject of scientific research for years. Modifications in human metabolism can alter the pattern of so-called volatile organic compounds (VOCs), which the body releases through breath, skin, and bodily fluids, among other channels.<\/p>\n The issue appears to be less about a single cancer molecule and more about complex chemical odour signatures. Dogs have around 300 million olfactory receptors, while humans have approximately five million. Their biological ability is currently difficult to replicate with technical sensors.<\/p>\n The research is not limited to cancer. Studies are also investigating the detection of covid-19, Parkinson’s disease, and other illnesses. According to Dognosis, numerous peer-reviewed studies on medical olfactory detection by dogs now exist.<\/p>\n Breathing into a mask for ten minutes<\/strong><\/p>\n The planned test is remarkably simple: a patient breathes into a cotton mask for about ten minutes. This is then hermetically sealed and sent to a laboratory where trained dogs examine the sample.<\/p>\n The patient has no contact with the animals. The goal is a standardized screening system that could also be organized via central laboratories.<\/p>\n An abnormal result would not necessarily be a definitive cancer diagnosis. Rather, it could prompt further investigations such as imaging, endoscopy, biopsy, or laboratory tests.<\/p>\n The algorithms are learning to understand each dog. Traditional detection dogs are often trained to exhibit specific indication behaviours. For example, the animal might sit in front of a positive sample or give another agreed-upon signal.<\/p>\n But dogs react differently. As Kulgod explains in the video, one dog might immediately run to the food dispenser because it expects its reward. Another might freeze, while yet another might scratch excitedly at the sample.<\/p>\n For Dognosis, this is not crucial. The dog does not need to learn to behave like a standardized measuring device \u2013 the machine should learn to understand the dog.<\/p>\n To achieve this, the company developed the \u201cDogSense<\/em>\u201d system. A non-invasive helmet and other sensors collect data including EEG, respiration, and behaviour. The \u201cDogOS<\/em>\u201d software processes this information using machine learning models.<\/p>\n The technology is designed to learn which biological reactions in a particular dog are associated with the perception of a disease odour. The dog provides the highly sensitive odour recognition; the technology translates its reaction into analysable data.<\/p>\n Phase II trial with 1,502 participants<\/strong><\/p>\n The study, published in 2026, is titled \u201cCanine Olfaction Combined with Bayesian Modelling for Multicancer Detection from Breath Samples: A Phase II Study in India<\/em>\u201d. Several Indian hospitals, Dognosis, and the British organization Medical Detection Dogs participated.<\/p>\n The study included 1,502 participants from six hospitals. Each breath sample was evaluated independently by at least three trained dogs. Subsequently, a statistical Bayesian model combined the results, taking into account, among other aspects, the previous performance of the dogs.<\/p>\n The results were remarkable: 90.8 percent sensitivity and 91.3 percent specificity.<\/strong> In simpler terms, this means that of 100 people with cancer, approximately 91 were correctly identified, and of 100 people without cancer, roughly 91 were correctly classified as negative. The AUC, a measure of the test’s ability to distinguish between cases, was 0.962.<\/p>\n According to the researchers, the method worked across seven major cancer groups and encompassed more than 20 cancer types. Particularly important is that its effectiveness remained high even in early stages of cancer.<\/p>\n This is precisely where the medical potential lies: many types of cancer can be treated more successfully if they are detected early. A cost-effective, non-invasive test that could detect several types of cancer simultaneously would therefore be of considerable benefit.<\/p>\n From sniffer dog to medical measuring system<\/strong><\/p>\n The fact that dogs can smell cancer is only half the problem. The other half is: how can this be turned into a reproducible medical procedure?<\/p>\n A hospital cannot base a diagnostic standard on the fact that a particular Labrador scratches today and a Beagle stands still for five seconds longer in front of a sample tomorrow. Medical tests need to be standardizable, reproducible, and quantifiable.<\/p>\n This is precisely where Dognosis’s true innovation lies. The company isn’t trying to build an artificial nose that smells better than a dog’s. It utilizes existing biological sensor technology and complements it with a digital measurement platform.<\/p>\n The dog provides the scent detection \u2013 sensors and algorithms are intended to turn this into standardized diagnostic information.<\/p>\n Promising, but not yet a replacement for established diagnostics<\/strong><\/p>\n As impressive as the results are, they do not mean that dogs and technology can already replace echography, colonoscopy, CT scans and so on.<\/p>\n The published study is a phase II trial. It shows that the method can distinguish between cancer and control samples under controlled conditions. The crucial next question is whether it also works in a real screening population and actually leads to better health outcomes.<\/p>\n That’s a significant difference. With a specificity of around 91 percent, numerous people in a population with a low cancer rate could receive an abnormal result even though they don’t have cancer. Therefore, the procedure would initially be, at best, a preliminary screening step, but not a diagnostic endpoint.<\/p>\n Furthermore, several authors of the study are employed by or own shares in Dognosis, and the research was funded by the company. These conflicts of interest are disclosed, but underscore the importance of independent confirmatory studies.<\/p>\n The step into practice<\/strong><\/p>\n Dognosis is currently training 15 dogs and collaborating with medical facilities in India. According to recent media reports, the company is aiming for a commercial launch in 2027 and is already in contact with regulatory authorities in various countries.<\/p>\n This could transform what has seemed like a rather curious scientific observation for decades into a scalable diagnostic system for the first time.<\/p>\n The real significance of this development may lie in its unusual approach: while researchers are developing increasingly complex technical sensors and blood tests, Dognosis utilizes a biological capability that already exists.<\/p>\n