Optical screening of diabetes mellitus using non-invasive Fourier-transform infrared spectroscopy technique for human...

The Fourier-transform mid-infrared spectroscopy (FTIR) technique has not been used for diabetes diagnosis so far in clinical practice. We attempted to predict non-invasively blood hemoglobin A1c (HbA1c) levels by FTIR to enable evaluation and screening for diabetes. Twenty eight patients from age 20s to 80s, 14 males and 14 females, with and without diabetes, were examined in hospital as a pilot study, and their biochemical data were analyzed with infrared (IR) spectral data of the lip surface by FTIR spectroscopy. Some IR peaks of lip surface had significant correlations with blood glucose and HbA1c levels. Among several peaks in the spectra of lip, peaks at around 1300-1400cm(-1) discriminated groups with higher or lower HbA1c levels, suggesting the observation of lip surface advanced glycation end-products with carboxymethyl group. The application of partial least squares (PLS) regression analysis to the correlation between HbA1c levels and lip FTIR spectra resulted in highly significant prediction of HbA1c values for the subjects. Comparison of some IR peaks with predicted HbA1c values served to remove false-negative data effectively. This study with FTIR technique proposed here was effective for screening of diabetes patients with higher HbA1c levels.