Abstract
OBJECTIVE: Accurate, easily accessible and economically viable cancer diagnostic tools are pivotal in improving the abysmal 5% survival rate of pancreatic cancer. METHODS: A novel, affordable, non-invasive diagnostic method has been developed by combining measurement precision of infrared spectroscopy with classification using machine learning tools. RESULTS: Diagnosis accuracy as high as 90% has been achieved. The study investigated urine and blood from pancreas cancer patients and healthy volunteers, and significantly improved accuracy by focusing on sweet-spots within blood plasma fractions containing molecules within a narrow range of molecular weights.