Diabetes is a chronic disease that affects millions of people every year worldwide. Patients with diabetes have high levels of glucose in their blood, since their bodies cannot produce or adequately use the insulin produced. Identification of diabetes is usually performed by tests of glucose in the blood by means of colorimetric reactions, which are time consuming and use a considerable amount of reagents. Attenuated total reflectance Fourier-transform infrared (ATR-FTIR) spectroscopy has been used in clinical research as a potential tool to obtain spectrochemical information of biological materials. The infrared spectra can be used as source of information for classiciation models and biomarker extraction by using specific computational tools. In this paper, a semi-portable Bruker Alpha ATR-FTIR was employed to analyse urine samples of 7 patients (3 normal, 2 diabetics and 2 pre-diabetics) in order to distinguish these three groups based on their spectrochemical information. Cross-validated principal component analysis, coupled with linear discriminant analysis was applied to the spectral dataset, resulting in 94% total accuracy. Sensitivities were observed to be 95%, 96% and 100% for normal, pre-diabetics and diabetics patients, respectively, with specificities of 93%, 91% and 100%. These findings show the potential of ATR-FTIR as a new possible tool for identification of diabetics in clinical environments, whereby the diagnosis can be performed in a rapid, non-invasive and automated way.
|