Receiver Operator Characteristic Analysis of Biomarkers Evaluation in Diagnostic Research

Receiver Operator Characteristic (ROC) analysis is the choice of method in evaluation of biomarkers in bioinformatics research. However, there is no single method and also no single accuracy index in evaluating diagnostic tools. This review provides an extensive illustration of different methods of ROC curve analysis that can be used in clinical practice of diagnostic studies. It includes their early use for rating data and the recent developments for quantitative data with a discussion of choice of model selection in parametric ROC analysis compared with non-parametric approach. The relevant methodological issues of these two alternative approaches have been discussed in terms of bias and sampling variability of Area under the curve (AUC) index that may influence on the performance of diagnostic tests. The methods were illustrated with two relevant clinical examples. The semi-parametric and parametric model of mixture of Gaussian is comparable with purely nonparametric approach. The choice between methods depends on practical conveniences unless the presence of severe departure from binormality. The recent new development and the gaps in knowledge concerning their behaviours in actual applications for medical researches and a guideline for future research have been discussed.

Abstract licence: CC BY-NC-ND

A person counts once a day per paper. We keep no cookie and no address: a reader is a hash that changes every day. Any month under five reads shows no number.