The test is straightforward: a person breathes normally into a cotton face mask for 10 minutes. The mask is sealed, stored, and later evaluated by trained detection dogs at a central laboratory. Each sample is assessed independently by at least three dogs, and their assessments are combined using an advanced Bayesian statistical model that weighs each dog's track record and the participant's background information. No blood is drawn, no scan is needed, and no fasting is required.
The study was conducted in collaboration with Medical Detection Dogs, a UK-based charity and world leader in canine bio-detection research that has previously demonstrated dogs can detect the odour of diseases including prostate and bladder cancer, Parkinson's disease, malaria, and COVID-19. Six hospitals in Karnataka, India participated (RadOn Cancer Centre, Karnataka Cancer Therapy and Research Institute, and Karnataka Medical College and Research Institute in Hubli; St John's Medical College, Aster CMI Hospital, and Narayana Health City in Bangalore), with 1,502 people in the final test group (283 with biopsy-confirmed cancer and 1,219 without). The dogs and model together correctly identified over 90% of cancers and over 91% of non-cancers across head and neck, breast, lung, gynecologic, upper and lower GI, and genitourinary cancers. Accuracy was just as high for early-stage disease (Stages I and II), where catching cancer early can dramatically improve survival.
"We've known for over two decades that dogs are capable of detecting multiple types of cancers with high accuracy," said Akash Kulgod, co-founder and CEO of Dognosis, the company behind the test. "The challenge has always been building a system around canine olfaction that is reproducible, scalable, and aimed at a clinical problem worth solving. Multi-cancer risk stratification from a single breath sample in countries like India is that problem, and this study shows that it can be done."
Dr. Basavaraj R. Patil, Chairman of the Karnataka Cancer Therapy and Research Institute (KCTRI) in Hubli and a co-author on the study, has practiced surgical oncology in the region for over three decades.
"We have initiated cancer screening through a dedicated van at our institute, which has increased outreach, but it is still a challenge to address the growing cancer incidence in India. A low-cost, repeatable, and non-invasive tool like this could be transformative, not only for early detection but also for monitoring whether treatment is working or detecting early signs of recurrence."
Claire Guest, co-founder of Medical Detection Dogs and a co-author on the study, said:
"We were delighted to collaborate with Dognosis to advise and support in the training of dogs to detect multiple cancers by odour. The results are staggering yet not surprising to us. Dogs have proved time and again that their ability to detect disease is superior to any other current options. We hope these results can be used to help advance and speed up diagnostic resources and screen many more people around the world."
Seven dogs took part in the study: four Beagles, one Labrador, one Labrador-Indie mix, and one Dutch Shepherd-Belgian Malinois mix, each trained over 10 weeks using reward-based methods. Analysis showed that fewer than 2% of variation in their responses was down to which individual dog was doing the assessment, meaning the system is consistent regardless of which dogs are on duty.
The study also showed early results from a next-generation version of the system that uses camera-based analysis of the dogs' sniffing behaviour to further refine accuracy. Dognosis is developing AI tools that read the dogs' behaviour and brain signals in real time, with the long-term goal of scaling the technology to screen millions of people a year. The publication is accompanied by an invited editorial and the authors have been invited to discuss the findings on the journal's podcast.
Dognosis is now preparing to evaluate the test in real-world programmes across multiple Indian states, with a U.S. study to follow.
No comments:
Post a Comment
Note: Only a member of this blog may post a comment.