Imagine a future where man's best friend joins forces with cutting-edge technology to sniff out undetected cancers before they have the chance to spread. With cancer ranking as one of the leading causes of death worldwide, the need for early detection continues to grow. In 2022 alone, doctors diagnosed nearly 20 million new cancer cases, and 9.7 million people lost their lives to these diseases. For decades, researchers have raced to uncover hidden signs of cancer during its early stages in the hope of discovering a method of detection that could dramatically increase survival rates. An unlikely duo—canines and artificial intelligence (AI)—may hold the key to this pursuit.

Above: A beagle on a new team of dogs being trained to sniff out breast, lung, colorectal, and prostate cancer. Image courtesy of SpotitEarly.

Doctors often recommend that individuals at risk for cancer undergo screening programs to detect the disease while still in its early stages. However, many current cancer screening methods have notable drawbacks. Some, like mammograms and CT scans, involve exposure to ionizing radiation, while others, such as colonoscopies, are invasive. Beyond the physical drawbacks of these methods, many people struggle to access these screening tools, particularly in underserved and rural areas. Combined with high costs and a lack of understanding about the procedures, these barriers result in low screening compliance, leaving many cancers undetected until it's too late.

Above: A CT scanner, a technology often used as a cancer screening method. Image courtesy of In Focus Radiology.

To address cancer screening hurdles, a company founded in 2020 called SpotitEarly has developed a simple, non-invasive, and self-administered screening method that detects cancer using exhaled human breath samples. This method relies on three main principles: the distinct molecular profile of cancer in breath samples, the odor-sensory abilities of canines to detect this molecular profile, and the use of AI to analyze canine responses to determine the presence of cancer. 

The first principle, the molecular profile of cancer found in our breath, is supported by evidence suggesting that tumor cells and their microenvironment produce a distinct volatile organic compound (VOC) pattern. Recently, a research team based in Israel assessed the ability of dogs to detect this molecular profile using biological samples such as sweat, blood, saliva, and breath. Although some studies have shown promising results with breath samples, notable differences remain in the performance of canines across studies. The specific chemical compounds detectable by canines are still not fully understood. Several factors, including training methodology and genetics, can also influence the performance of canine scent detection. Lastly, AI analysis of the data relies on AI tools to drive automation, enhance precision, and improve accuracy. 

Above: Illustration representing how volatile organic compounds (VOCs) in breath samples contain unique cancer odor signatures. Image courtesy of SpotitEarly.

The research team conducted a double-blind study—in which neither researchers nor the trained canines knew of a patient’s cancer status—to evaluate two critical measures of the diagnostic accuracy of these methods: sensitivity (the ability to correctly identify people with cancer) and specificity (the ability to correctly identify people without cancer). The study focused on four major cancers—breast, lung, prostate, and colorectal cancer—that collectively represent nearly half of all cancer diagnoses. Researchers collected breath samples from approximately 1,400 participants and selected six Labrador Retrievers to train to mark samples as positive for cancer through a distinct behavioral cue: sitting beside the sample immediately after sniffing. 

The canine testing protocol was rigorous. Each dog entered the detection room individually and examined the samples multiple times. While the dogs performed their assessment, an AI system simultaneously analyzed both their trained response (sitting) and various unconscious behavioral and physiological signals captured by the room's sensors and cameras. This dual-analysis approach generated a binary cancer prediction (positive or negative) for each sample. To validate the results, researchers compared these predictions against traditional diagnostic methods, including standard cancer screenings and biopsy results.

Above: An illustration of SpotitEarly’s testing and control (monitoring) room. Image courtesy of Half et al., 2024. 

Researchers found that when testing for lung, breast, colorectal, and prostate tumors, the system correctly identified 93.9% of people who had cancer (sensitivity) while accurately ruling out 94.3% of those who did not (specificity). More notably, the method maintained its exceptional performance even for early-stage cancers, successfully detecting 94.8% of cases. This success rate sets it apart from other screening tests that struggle to maintain high sensitivity for early-stage cancer detection without compromising specificity.

SpotitEarly's screening method matched the sensitivity of gold-standard tests across all targeted cancers. The key advantage, however, is that SpotitEarly combines screening for these four common cancers into a single, non-invasive test, making it a promising complement to existing screening protocols. The method's simplicity and accessibility could transform early cancer detection. Since breath samples can be collected even in non-clinical settings, this method could significantly boost compliance with cancer screening among the general population.

Perhaps even more fascinating, this new screening method showed unexpected potential beyond its original scope. Despite the canines being trained for only four main cancer types, the method yielded high-performance results for an additional fourteen types of cancer. While these applications are still exploratory, they suggest that different cancers may share similar VOC signatures in their unique molecular profiles. This discovery opens up exciting possibilities for expanding SpotitEarly's capabilities to screen for an even broader range of cancers in the future. SpotitEarly is currently planning a larger clinical trial in the United States and aims to report early results in 2026.