This study aims to differentiate early between patients with osteoporosis (OP) and low bone density (LBD) using Fourier transform infrared (FTIR) spectroscopy combined with the partial least squares‐support vector machine (PLS‐SVM) algorithm. The study included 78 patients aged 50–90 years, categorized into three groups based on bone mineral density (BMD): normal (N group), LBD group, and OP group. FTIR spectra of serum samples were acquired using potassium bromide pellets of uniform thickness, and PLS‐SVM was employed for dimensionality reduction and classification. The results demonstrated an accuracy of 95%, 85%, and 90% for the N, LBD, and OP groups, respectively, in the test set. Second‐derivative spectral analysis revealed that the spectral intensities at the specific wavenumbers of 2925, 2860, 1651, 1610, 1523, 1505, 1481, 1421, 1399, 1361, and 1139 cm −1 were lower in OP and LBD patients compared to the normal group, indicating differences in lipids, proteins, and amino acids. This study offers a promising new, noninvasive, cost‐effective, and sensitive alternative for the early diagnosis of OP.
Yang et al. (Thu,) studied this question.