Abstract: Skin autofluorescence arises from the superposition of various endogenous fluorophores in tissues upon excitation, serving as an indicator of related disease states. However, spectral overlap among endogenous fluorophores complicates the detection of specific targets, so selecting optimal detection wavelengths can enhance the identification effectiveness of these targets. Advanced glycation end products (AGEs), reduced niacinamide adenine dinucleotide (NADH), flavin adenine dinucleotide (FAD) and collagen (COL) were selected as experimental subjects in this work. Firstly, a database of overlapping three-dimensional fluorescence spectra of the four fluorescent substances were constructed. Then, the characteristic regions of the four fluorescent substances were identified using the region integration method, and 178 characteristic wavelengths were selected through the elimination of uninformative variables. Finally, a calibration curve was established based on multiple linear regression, and the recovery rates of each fluorescent substance were calculated. The results showed that the recovery rates of AGEs, NADH, FAD, and COL were 70%–110%, 80%–110%, 70%–130%, and 50%–190%, respectively, and the predictive accuracy for AGEs, NADH, and FAD of the model based on 178 characteristic wavelengths was acceptable, achieving highly selective screening of detection wavelengths for the three fluorescent substances in overlapping spectra.

Key words: biomedical optics, characteristic fluorescence, three-dimensional fluorescence spectrum, advanced glycation end products