Diagnosis of hyperthyroidism and hypothyroidism serum by SVM-based Raman spectroscopy

Abstract

The Raman spectra of 29 people with normal thyroid function, 38 cases of hyperthyroidism and 32 cases of hypothyroidism were obtained with a portable Raman spectrometer with a confocal optical path design. The peaks of the Raman spectra of the three groups were analyzed. Intensity analysis indicates that the Raman spectral intensities of the three groups have a significant difference in certain areas, namely at 1002 cm−1, 1145 cm−1 and 1511 cm−1. The partial least squares (PLS) algorithm, combined with a support vector machine classification method, was used to realize the differential diagnosis of hyperthyroidism and hypothyroidism at the molecular level. The PLS data model analysis shows that the Raman spectra have significant differences in the principal components from PLS-1 to PLS-8, and, according to the 3D scattergram, healthy people, patients with hyperthyroidism and hypothyroidism serum samples can be distinguished. The support vector classifier (SVC) was used for data classification. The specificity of diagnosis is 88.8%, the sensitivity is 100%, and the total discriminant accuracy is 96.66%. Studies have shown that Raman spectroscopy and the PLS-SVC classification method are expected to be auxiliary tools for clinical diagnoses of hyperthyroidism and hypothyroidism.