Speaker
Description
Cancer remains the second leading cause of death across EU countries, with ovarian cancer being one of the deadliest gynecological malignancies. Its asymptomatic onset and delayed diagnosis contribute to poor survival outcomes—only 4% at stage IV compared to 90% at stage I. Existing diagnostic tools lack the sensitivity, specificity, and scalability needed for routine early screening. There is an urgent need for minimally invasive, accurate, and cost-effective diagnostic strategies to enhance early detection and reduce cancer burden.
We present a novel, non-invasive approach leveraging volatile organic compound (VOC) profiling from blood plasma using a 32-element metal oxide semiconductor-based electronic nose, integrated with advanced machine learning algorithms. Each 1 ml plasma sample was measured over 10 minutes after standardized thawing procedures. Custom feature extraction and sensor selection enabled the training of a boosting-based ensemble model, achieving 97% sensitivity and 97% specificity in distinguishing ovarian cancer from healthy controls. A majority-vote classification framework yielded 100% patient-level diagnostic accuracy.
Beyond binary classification, the system demonstrated the capacity to identify cancer stage and distinguish between ovarian and endometrial cancer. This compact, rapid, and portable diagnostic platform offers significant potential for scalable, real-time early cancer detection and screening, especially in low-resource settings.
By integrating data-driven methodologies with gas sensor technology, this work challenges conventional approaches in cancer diagnostics—pushing the boundaries of what is feasible in the clinical setting. The results show the power of interdisciplinary applied physics in transforming biomedical diagnostics and underscore the importance of embracing diverse trajectories in both research and healthcare innovation.