An expert system approach for the objective interpretation of serum tumour marker levels.

The aim of our work was to study specific aspects of statistical Weinstein-Bayes's analysis in order to find the most suitable program for presenting our results and for obtaining information which could be used as a guide in the clinical decision-making process. Our efforts resulted in a computer processing system based on modified Weinstein-Bayes receiver operating characteristics (ROC) analysis. The data analysis provides numeric values of sensitivity, specificity, positive predictive value and negative predictive values as output tumour marker databases, which would serve as graphic input. It becomes possible to plot the ROC curve as a graphical distribution of tumour marker levels. The computer program also allows the estimation of the potential utility of the serum tumour marker measured in patients with various diseases, simultaneous determination of serum tumour markers in diseased patients, and the real value of each serum tumour marker level in diseased patients.