Near-infrared spectroscopic applications for diagnosis of endometrial carcinoma

NIR spectra of 77 endometrium sections (malignant, hyperplasia, and normal samples) are collected. Partial least squares discriminant analysis (PLS-DA) and fuzzy rule-building expert systems (FuRES) are used for classification based on the NIR spectral data. The classification ability of two classifiers is evaluated by using ten bootstraps and five Latin partitions. The results indicate that the classification ability of FuRES is better than that of PLS-DA. The sensitivity, specificity, and accuracy obtained from FuRES for malignant endometrium diagnosis are 90.0±0.7, 95.0±0.8, and 93.1±0.8%, respectively. The results demonstrate that NIR spectroscopy combined with the FuRES technique is promising for the classification of endometrial specimens and for practical diagnostic applications.