Noninvasive detection of diabetes mellitus based on skin fluorescence and diffuse reflectance spectroscopy

• Published in: Journal of biophotonics Vol. 17; no. 1; pp. e202300098 - n/a
• Main Authors: Zhang, Yuanzhi, Liu, Yong, Zhu, Guoqing, Wang, Quanfu, Ni, Jingshu, Liu, Lin, Zhang, Jian, Zhang, Junqing, Li, Zhongsheng, Wang, Xia, Huang, Yao, Dong, Meili, Zhang, Yang, Wang, Yikun
• Format: Journal Article
• Language: English
• Published: Weinheim WILEY‐VCH Verlag GmbH & Co. KGaA 01.01.2024; Wiley Subscription Services, Inc
• Subjects: Accuracy; Algorithms; Chemical compounds; Cohort Studies; Diabetes; Diabetes mellitus; Diabetes Mellitus - diagnostic imaging; diffuse reflectance; Diffuse reflectance spectroscopy; Emission measurements; Excitation spectra; Fluorescence; Fluorophores; Humans; Light sources; Mass Screening; Mathematical models; Modelling; multi‐peaks Gaussian fitting; Parameters; Physiology; Population studies; Reflectance; Scattering coefficient; Skin; Skin - diagnostic imaging; Spectrometry, Fluorescence - methods; Spectroscopy; Spectrum Analysis; spectrum recovery; tissue fluorescence spectrum; diabetes mellitus; multi-peaks Gaussian fitting; spectrum recovery; tissue fluorescence spectrum; diffuse reflectance
• ISSN: 1864-063X; 1864-0648; 1864-0648
• DOI: 10.1002/jbio.202300098

There is an urgent need for a mass population screening tool for diabetes. Skin tissue contains a large number of endogenous fluorophores and physiological parameter markers related to diabetes. We built an excitation‐emission spectrum measurement system with the excited light sources of 365, 395, 415, 430, and 455 nm to extract skin characteristics. The modeling experiment was carried out to design and verify the accuracy of the recovery of tissue intrinsic discrete three‐dimensional fluorescence spectrum. Blood oxygen modeling experiment results indicated the accuracy of the physiological parameter extraction algorithm based on the diffuse reflectance spectrum. A community population cohort study was carried out. The tissue‐reduced scattering coefficient and scattering power of the diabetes were significantly higher than normal control groups. The Gaussian multi‐peak fitting was performed on each excitation‐emission spectrum of the subject. A total of 63 fluorescence features containing information such as Gaussian spectral curve intensity, central wavelength position, and variance were obtained from each person. Logistic regression was used to construct the diabetes screening model. The results showed that the area under the receiver operating characteristic curve of the model for predicting diabetes was 0.816, indicating a high diagnostic value. As a rapid and non‐invasive detection method, it is expected to have high clinical value. The system for measuring tissue excitation‐emission spectra and diffuse reflectance spectra has been designed and manufactured. It is capable of collecting these spectra from subjects' skin. With the integration of a built‐in diabetes recognition model, the system is expected to enable fast screening of high‐risk populations for diabetes, thereby contributing to diabetes prevention and control efforts.