A machine-learning–enabled “electronic nose” capable of identifying ovarian cancer–associated signatures from blood plasma samples has been described in a study published in Advanced Intelligent Systems.
Early detection of ovarian cancer remains challenging because symptoms are often vague and current diagnostic approaches lack sufficient sensitivity for routine screening. As a result, many cases are diagnosed at advanced stages.
In the study, researchers developed a diagnostic platform combining a 32-sensor electronic nose with machine-learning algorithms designed to analyse volatile organic compounds released from blood plasma samples. Rather than targeting a specific biomarker, the system evaluates complex chemical patterns associated with disease.
CLINICAL SUMMARY
What was examined
A study evaluated a machine-learning–enabled electronic nose that analyses volatile organic compounds released from blood plasma to identify ovarian cancer–associated signatures.
Key findings
-
A 32-sensor electronic nose combined with machine-learning algorithms analysed volatile organic compounds from plasma samples.
-
The system demonstrated high diagnostic performance in distinguishing ovarian cancer samples from healthy controls.
-
The model was also able to distinguish ovarian cancer from endometrial cancer samples and showed potential for identifying disease stage.
Clinical implications
-
Electronic-nose technology may represent a potential non-invasive approach to cancer detection based on chemical signatures rather than individual biomarkers.
-
Larger prospective studies will be required to determine clinical utility and any potential role in screening or early diagnosis.
Using plasma samples from a biobank, the investigators trained and validated the model to distinguish ovarian cancer samples from healthy controls. In testing, the system demonstrated high diagnostic performance, with sensitivity and specificity reported in the high-90% range. Patient-level classification reached 100% accuracy using a majority-vote approach within the study dataset.
The analytical framework was also able to distinguish ovarian cancer from endometrial cancer samples and showed potential for identifying disease stage.
Featured Event
Electronic nose technologies have been explored for several decades, but advances in machine learning are improving the ability of sensor arrays to interpret complex chemical signatures. The authors note that combining these technologies may support future approaches to cancer detection that do not rely on predefined biomarkers.
However, further validation in larger clinical cohorts will be required before the approach can be considered for clinical implementation or potential screening applications.
Paper: Shtepliuk Ivan, Meng Lingyin, Borgfeldt Christer, Eriksson Jens, Puglisi Donatella, Adv. Intell. Syst. 2025; 000, e202500838. Access online here.
