Disease pattern recognition in FT-IR spectra of human sera

We observed differences between the mid-infrared spectra of sera originating from healthy volunteers and from patients with diabetes mellitus or rheumatoid arthritis. These differences were found to be significant in terms of the Fisher criterion, the t-test, and the Kolmogorov-Smirnov test. The significance allows for a classification of the spectra and a probability ("DPR-score") of belonging to the class "healthy" can be computed. In comparing the samples from 80 diabetes patients with samples from 40 healthy volunteers we are able to achieve a sensitivity and a specificity of 80% and above. The DPR-score correlates better with the actual status of health than the glucose concentration alone. In a study on rheumatoid arthritis we compared the spectral signatures of sera taken from 188 rheumatoid arthritis patients and sera from 196 healthy volunteers. By applying linear discriminant analysis to 2/3 of the samples we are able to classify the remaining third of the samples (independent validation) with a sensitivity of 84% and a specificity of 88%.