Classification Algorithms for SIFT-MS Medical Diagnosis

Selected ion flow tube-mass spectrometry (SIFT-MS) is an analytical technique for the real-time quantification of trace gases in air or breath samples. The SIFT-MS system can potentially offer unique capability in the early and rapid detection of a wide variety of diseases, infectious bacteria and patient conditions, by using a classifier to differentiate between control and test groups. By identifying which masses and volatile organic compounds(VOCs) contribute most strongly towards a successful classification, biomarkers for a particular disease state may be discovered. A classification method is presented and validated in a simple study in which saturated nitrogen in tedlar bags was differentiated from dry nitrogen in tedlar bags. Several biomarkers were identified, with the most reliable being N₂H⁺.H₂O, and isotopes and water clusters of H₃O⁺, as expected. The classifier was then applied in a clinical setting to differentiate between patient breath samples after one and four hours of dialysis treatment. Biomarkers for classification were ammonia, acetaldehyde, ethanol, isoprene and acetone. The model classifies significantly better than random, with an ROC area of 0.89.