Dogs + AI Sniff Cancer — Doctors Stunned

Lab technician using an analyzer with a pipette — Photo: Bolyuk Studio / Shutterstock

A Bengaluru startup says trained dogs plus artificial intelligence spotted cancer from breath samples with about 90% accuracy across seven cancers.

Story Snapshot

Dognosis reports 90%+ sensitivity and specificity in a Phase 2 trial of 1,502 people.
The method pairs trained dogs with sensors and machine learning to read breath samples.
The company says results were strong even in early-stage cancers and across types.
Leaders pitch a low-cost, non-invasive test that could expand screening access.

What Dognosis Reported From Its Phase 2 Trial

Dognosis announced Phase 2 results that it says reached over 90 percent sensitivity and specificity, with an area under the curve of 0.962, in a study of 1,502 participants. The company stated that performance held across seven major cancers and at Stage 1 and Stage 2. Dognosis framed the study as the largest of its kind and linked it to publication in the Journal of Clinical Oncology. These claims come from the company’s post and newsroom summary.

Dognosis’s newsroom page repeats the core finding: a breath-based test, powered by trained dogs, detected cancer with over 90 percent accuracy across several types. It highlights early detection as a key benefit and points to potential use for screening. The company positions the work as part of a growing field focused on volatile organic compounds in breath that can signal disease. The newsroom post attributes the trial to Kulgod and colleagues in 2026.

How the Breath Test and Dog-AI System Works

The startup’s workflow starts with a person breathing into a special mask for about ten minutes. The mask traps volatile compounds in the breath. Trained dogs then sniff the sample. A headset and other sensors record the dogs’ brain and body signals as they react. Machine-learning software translates those signals into a yes or no output for possible cancer. The company says this mix can scale the dogs’ skill into a repeatable screening tool.

Company materials describe the approach as multi-layered: dog olfaction, a brain-computer interface, and analytics. The goal is to turn a dog’s detection into quantitative data patterns that software can compare at large scale. Leaders say the design could deliver a low-cost, non-invasive test that reaches people who lack access to major hospitals or scans. That promise has drawn attention in India’s startup press and health coverage highlighting access and cost pressures.

Why This Could Matter for Patients and Health Systems

Early cancer detection often raises survival odds and can lower treatment costs. Many people skip or delay screening because of price, travel, or fear of invasive tests. A simple breath test done at home or in a clinic could close that gap. Dognosis promotes the test as “ultra-affordable” and easy to use. If real-world performance matches trial figures, systems could triage patients faster and ease burdens on imaging and specialty clinics, especially in lower-resource areas.

The company also ties its work to a broader push to “teach” artificial intelligence to smell. Past studies have shown dogs can pick up disease scents, and some teams have tried to train software on those patterns. The National Institutes of Health has highlighted efforts to learn from canine detection to build machine models. Dognosis aims to merge both: keep the dogs in the loop now while training algorithms that may one day stand alone at scale.

What We Still Do Not Know From Public Materials

Public posts do not break down results by each cancer type, exact false positive rates in healthy people, or how the test performed in common real-world settings. The company’s message stresses strong accuracy and early-stage detection, but external validation details are limited in available summaries. As with many new screening tools, health systems will watch how the method works outside study settings, what the cost per test is, and how it integrates with current care paths.