拉曼光谱技术被认为是最有希望实现人体血糖无创检测的技术之一。针对无创血糖检测技术存在信噪比低、校正模型建立困难的问题，项目提出空间偏移拉曼光谱（SORS）和深度学习算法相结合的无创血糖检测方法，提高检测的系统信噪比和模型预测精度，以求血糖预测精度达到5mg/dL的临床应用要求。拟开展SORS无创血糖检测的理论模型研究；高信噪比SORS光谱检测系统设计与优化；子空间特征提取结合深度学习的血糖分析模型构建等研究工作。其创新性为：通过光纤源探距离（ΔS）优化，实现对目标组织的血糖信号采集，减少其他组织层的信号干扰，提高系统的信噪比；通过“球透镜+斜面光纤”光纤探头结构优化，增强入射光光强，增大收光范围，提高系统的信噪比；通过随机子空间法，实现对拉曼光谱更深层次血糖特征信息的挖掘，提高血糖预测精度。项目最终目标是研发一种人体血糖无创高精度检测技术，实现无创血糖临床评估及监测。

Raman spectroscopy is considered one of the most promising technologies for non-invasive detection of human blood glucose. To resolve the issues of current non-invasive blood glucose detection technologies, such as low signal-to-noise ratio and difficulty in establishing the calibration model, this project has proposed a non-invasive blood glucose detection method that combines spatially-offset Raman spectroscopy (SORS) with deep learning algorithms. This novel method can improve the system signal-to-noise ratio and the accuracy of the detection prediction model, so that the clinically-required blood glucose detection accuracy of 5 mg/dL can be satisfied. The project also proposes studying the theoretical model of the SORS non-invasive blood glucose detection technology, design, and optimize an SORS detection system with a high signal-to-noise ratio and establish a blood glucose analysis model that integrates subspace feature extraction with deep learning. This project is innovative because by optimizing the fiber source-detector distance (ΔS), the target tissue’s blood glucose signal is acquired. This process is completed at an escalated signal-to-noise ratio by reducing the signal interference from other tissue layers. By utilizing an optimized “ball lens + bevel fiber” coupled fiber probe, the incident light intensity is enhanced, the light absorption range is increased, and the diffusion photon absorption efficiency is improved. Finally, by investigating deeper blood glucose characteristic information in the Raman spectrum through random subspace, a highly accurate blood glucose calibration model can be established. The ultimate goal of the project is to develop a non-invasive high-precision blood glucose detection technology to achieve clinical evaluation and monitoring of non-invasive blood glucose.