A non-invasive diagnostic system engineered at the University of Pennsylvania Perelman School of Medicine combines artificial intelligence and machine learning to detect cancer.

The electronic nose, or e-nose, identifies hard-to-detect forms of the disease, such as pancreatic cancer and ovarian cancer, with up to 95% accuracy. The odor-based diagnostic relies on nanosensors to sniff out volatile organic compound (VOC) vapors emanating from blood plasma samples. The researchers previously demonstrated that VOCs released from plasma samples of ovarian cancer patients are distinct from those released by patients with benign tumors.

When administered to control subjects and cancer patients, the 20-minute test system effectively distinguished VOCs from ovarian cancer with 95% accuracy and pancreatic cancer with 90% accuracy, and correctly identified all patients with early-stage cancers.

The research results will be presented on June 4, 2021, at the annual American Society of Clinical Oncology meeting.