Objective: Evaluation of the potential of combined multivariate chemometric methods for seeking markers of lung cancer.
Methods: Statistical methods such as Mann-Whitney U test, discriminant function analysis (DFA), factor analysis (FA) and artificial neural network (ANN) were applied to evaluate the obtained data from GC/MS analysis of exhaled breath.
Results: The total number of compounds identified by GC/MS in human breath was equal to 88. The statistical analysis indicates seven analytes which have the highest discriminatory power. Cross validation of the obtained model shows that the sensitivity was 80% and the specificity was 91.23%, while for the test group the sensitivity and specificity were both 86.36%.
Conclusion: The application of combined statistical methods allowed to reduce the number of compounds to significant ones and indicates them as markers of lung cancer.
Keywords: Artificial neural network; Breath samples; Discriminant function analysis; Lung cancer; Statistical analysis; Volatile organic compounds.
Copyright © 2019 Elsevier B.V. All rights reserved.