Application of principal component regression in Mach–Zehnder interferometer optical fiber sensors in reflection mode for acetone detection as biomarker of diabetes mellitus

Acetone detection in human breath is important since it is a diabetes mellitus biomarker frequently found in concentrations in the order of 1.8 ppm for human patients. On the other hand the use of new data analysis techniques in Mach–Zehnder interferometer sensor systems have been increasing since they allow to explore the behavior of all range of transmission or reflection spectrum, providing more information that is not evident or sometimes perceptible on such spectra, hence other kind of techniques for characteristics extraction and data analysis such as principal component analysis, and principal components regression are used. They are methods that reduce dimensionality and can find some relationship or pattern from the data under analysis. In this work we show the performance of these techniques on the measured responses of three optical fiber sensors that are based on a Mach–Zehnder interferometer in reflection mode. The limit of detection found using principal component regression was approximately 1.5 ppm for acetone, which could be used to detect diabetes mellitus using breath analysis.