Researchers at Linköping University have developed an AI-enhanced electronic nose capable of detecting early signs of ovarian cancer in blood samples with high accuracy. The pilot study demonstrates how established sensor technology combined with machine learning could open new avenues for cancer screening.
Ovarian cancer is often diagnosed at an advanced stage, as early symptoms are vague and easily mistaken for more common conditions. Globally, around 325,000 new cases and more than 200,000 deaths were reported in 2022, with incidence expected to rise significantly in the coming decades. Unlike breast cancer, there is currently no reliable screening method for early-stage ovarian cancer.
Mimicking the human sense of smell
The research team set out to replicate the mammalian sense of smell using artificial intelligence. Their prototype electronic nose contains 32 commercially available sensors that respond to volatile organic compounds emitted from blood plasma samples. Different cancers release distinct patterns of volatile substances, effectively giving each disease a unique “odor signature.”
Rather than targeting a single biomarker, the system analyses complex patterns of volatile compounds. Machine-learning models, trained on biobank samples with known diagnoses, identify disease-specific signatures. In the pilot study, the algorithm distinguished ovarian cancer from endometrial cancer and healthy controls with an accuracy of 97 percent.
According to associate professor Donatella Puglisi, the approach could facilitate more accessible screening: “If screening were more accessible, both in terms of cost and location, it would be possible to improve early diagnosis.” The study was recently published in Advanced Intelligent System.
Toward scalable, low-cost screening
Traditional blood-based cancer tests often rely on individual biomarkers and can be slow or lack sensitivity in early disease stages. By contrast, the electronic nose test takes approximately ten minutes and provides a clear result, according to Jens Eriksson, associate professor at Linköping University and CTO of VOC Diagnostics AB, which is developing the technology. “It’s a simple test that takes 10 minutes and gives a clear result. Our method can test many people at a low cost and is much more accurate than what’s on the market today,” Eriksson stated.
While the current study focuses on ovarian cancer detection, researchers believe the technology could be adapted to identify multiple cancer types. Clinical validation is still required, but the team aims to integrate the tool into cancer screening programmes within the next three years, marking a potential step forward in AI-driven, non-invasive diagnostics.
