Use of multidimensional data analysis for prediction of lung malignity

Diagnosis of lung malignity can be predicted or confirmed not only according to the values of appropriate laboratory tests but also using multidimensional statistical analysis, which uses simultaneously all performed tests in the form of their optimal combination. The developed new way of diagnosis prediction is applied here to the results of laboratory analysis of lung tumor markers in serum as well as pleural effusion (exudate). Four laboratory tests were used and investigated in detail: carcinoembryonic antigen, CEA, in serum as well as in pleural exudate, and cytokeratin 19 fragment, CYFRA, in serum and exudate, as well. Each test represents one dimension in the investigated biomedical problem from the statistical point of view. Joint utilization of the performed laboratory tests is based on their optimized combination into a new statistical variable using a selected chemometric principle (principal component, discriminant function, or logit in logistic regression). This approach results in enhancement of diagnostic effectiveness applied for the specified purpose.