Abstract: In order to screen potential diabetes patients in high-risk groups, a non-invasive screening method for diabetes in high-risk groups based on optical coherence tomography (OCT) is proposed, and the patients could be early treated and reduce complications. This method first divided the 3-D OCT image into many small regions. Due to the different correlations between blood glucose changes and scattering coefficients in different regions, the principal component analysis was used to transform the scattering coefficients of these blood glucose sensitive regions into a comprehensive indicator S. Then S was normalized at different times of 0 min~120 min to obtain the relative quantity S_i(i=0, 1, 2, …, 120), with a relative quantity of S₁₂₀ at 120 min. According to the experience value of normalized blood glucose of healthy subjects, moderate diabetes, and severe diabetes in 0 min~120 min, the judgment threshold 1 and threshold 2 is 0.47 and 0.78, respectively. If 0 < S₁₂₀ < 0.47, 0.47 < S₁₂₀ < 0.78, and 0.78 < S₁₂₀ < 1, the subject is healthy, moderate diabetes and severe diabetes, respectively. This method is high practicability and has great significance for screening diabetes in high-risk groups with optical non-invasive methods.