Chemometric approach for improving VCSEL-based glucose predictions

Sahba Talebi Fard, Lukas Chrostowski, Ezra Kwok, Markus Christian Amann

Research output: Contribution to journalArticlepeer-review

2 Scopus citations


Optical methods are one of the painless and promising techniques that can be used for blood glucose predictions for diabetes patients. The use of thermally tunable vertical cavity surface-emitting lasers (VCSELs) as the light source to obtain blood absorption spectra, along with the multivariate technique partial least squares for analysis and glucose estimation, has been demonstrated. With further improvements by using data preprocessing and two VCSELs, we have achieved a clinically acceptable level in the physiological range in buffered solutions. The results of previous experiments conducted using white light showed that increasing the number of wavelength intervals used in the analysis improves the accuracy of prediction. The average prediction error, using absorption spectra from one VCSEL in aqueous solution, is about 1.2 mM. This error is reduced to 0.8 mM using absorption spectra from two VCSELs. This result confirms that increasing the number of VCSELs improves the accuracy of prediction.

Original languageEnglish
Article number5268215
Pages (from-to)578-585
Number of pages8
JournalIEEE Transactions on Biomedical Engineering
Issue number3
StatePublished - Mar 2010


  • Chemometrics for glucose prediction
  • Optical glucose monitoring
  • Real-time glucose monitoring


Dive into the research topics of 'Chemometric approach for improving VCSEL-based glucose predictions'. Together they form a unique fingerprint.

Cite this