Accurate methods of rapid medical diagnostics would obtain recognition among clinicians. In this paper, we present a ROC (receiver operating characteristic) analysis based approach to investigate intrinsic fluorescence spectra of medical samples. The approach provides researchers with capabilities for both spectroscopic feature selection and classification tasks. The method is illustrated using data obtained from samples of uterine cervix's tissues (normal, inflammatory, and precancerous conditions), but it is applicable for data from other types of medical specimens too. The results indicate the possibility to identify spectral ranges at which differences between various medical conditions are most evident as well as potential to apply the method for inflammation (cervicitis) vs. other medical conditions' diagnostics.