A computerized classification technique for screening for the presence of breath biomarkers in lung cancer.

Abstract

Abstract A simple computer-based screening technique has been developed for classifying human expired air components into 16 chemical classes, based on empirical formulas. The sort procedure was developed to simplify the screening of the composition of expired air samples by sorting all components into chemical classes and classifying components at the greater than 75% and greater than 90% occurrence levels. Both occurrence-rate components are then evaluated as diagnostic markers in a discriminant function model for their ability to detect lung cancer. Of the 386 components detected in the gas chromatography/mass spectrometry (GC/MS) data files, 45 components were present at the greater than 75% occurrence level and 28 components at the greater than 90% occurrence level. Thus, this preliminary sort routine, performed by using a simple macro program installed into a standard personal-computer spread-sheet, greatly reduces the amount of data required for statistical treatment. Such a sort routine can also be applied as easily to other complex GC/MS data files for the purpose of data reduction.