AUTHOR=Li Yuanpeng , Guo Liu , Li Li , Yang Chuanmei , Guang Peiwen , Huang Furong , Chen Zhenqiang , Wang Lihu , Hu Junhui TITLE=Early Diagnosis of Type 2 Diabetes Based on Near-Infrared Spectroscopy Combined With Machine Learning and Aquaphotomics JOURNAL=Frontiers in Chemistry VOLUME=Volume 8 - 2020 YEAR=2020 URL=https://www.frontiersin.org/journals/chemistry/articles/10.3389/fchem.2020.580489 DOI=10.3389/fchem.2020.580489 ISSN=2296-2646 ABSTRACT=Early diagnosis of diabetes is an important way to reduce the incidence and mortality of diabetes . The feasibility of early diagnosis of diabetes was studied via near-infrared spectra (NIRS) combined with support vector machine (SVM) and aquaphotomics. Firstly, the NIRS of whole blood samples from the population of health, pre-diabetes and diabetes were obtained. The spectral data of the whole spectra visible and near-infrared region (400-2500nm) were used as the research object of qualitative analysis. Secondly, several preprocessing steps including multiple scattering correction (MSC), variable standardization (SNV), first derivative and second derivative, were performed, and the best pretreatment method was selected. Lastly, early diagnosis of diabetes models were established by using SVM. Furthermore, fingerprint region of water(1300-1600 nm) were used as the research object for aquaphotomics model, and the aquagram of health group, pre-diabetes group and diabetes group was draw by using 12 water absorption pattern for early diagnosis of diabetes.The results of SVM show that the highest accuracy is 97.22% when the pretreatment method of first derivative was used, and the best model parameters are c=18.76 and g=0.008583.The results of aquaphotomics show that the obvious differences were observed in 1400-1500 nm region, and the number of hydrogen bonds in water species (1408 nm, 1416 nm, 1462 nm and1522 nm) is obviously correlated with the occurrence and development of diabetes: the number of hydrogen bonds is the smallest in the healthy group and the largest in the diabetes group. The reason may be caused by the water environment of blood has changed obviously with the aggravation of blood glucose metabolic dysfunction, and the number of hydrogen bonds could be used as a biomarkers for early diagnosis of diabetes. Result show that it is effective and feasible to establish the early diagnosis model of diabetes via NIR combined with SVM and aquaphotomics, which provides an accurate and rapid method to realize the early diagnosis of diabetes.