Presentation Information
[9p-P01-7]Non-invasive blood component analysis by mid-infrared photoacoustic spectroscopy– Analysis of the type 1 diabetic spectra and its application to healthy subjects –
〇Kosei Sasaki1, Saiko Kino1, Yuji Matsuura1 (1.Tohoku Univ., Graduate School of Engineering)
Keywords:
Non-invasive blood glucose monitoring
The optical absorption spectra of individuals with type 1 diabetes were analyzed to improve a predictive model of blood glucose levels in healthy subjects. Wavelengths around 1085 cm-1 were identified as important for glucose estimation and were subsequently applied to the analysis of healthy individuals. To compensate for the limited data from healthy subjects, pseudo-spectra were generated to estimate blood glucose levels. As a result, partial least squares regression (PLSR) analysis yielded a coefficient of determination (R2) of 0.56, which was an improvement over the R2 of 0.44 obtained without wavelength selection. These findings suggest that spectral information derived from diabetic patients can also be useful for blood glucose estimation in healthy subjects.