A method for identifying edible oils based on sample fingerprinting library of low-field NMR signal

Low-field Nuclear Magnetic Resonance (LF-NMR) has been used to determine the quality of edible oils for decades of years [1] [2] . However, the resolution is not very high. In this paper, a method based on pre-created sample fingerprinting libraries is presented to analyze the LF-NMR signal of edible oils, in order to improve the resolution of LF-NMR in the identification of edible oils quality.
Firstly, several kinds of edible oils (of which illegal cooking oils are provided by Jiangsu Provincial Center for Disease Control and Prevention, and the other oils are purchased on JD Online Shopping Mall) are measured using a LF-NMR instrument (Bruker the minispec mq60) to obtain the free induction decay (FID) curves.
Then, four different sample fingerprinting libraries are built according to above FID curves. The four fingerprinting libraries are respectively: FID fingerprinting library built using the original FID curves as criterion, T₂ fingerprinting library built using exponential fitting on FID curves as criterion, T₂ spectrum fingerprinting library built using inverse Laplace transformation (ILT) on FID curves, and peak area fingerprinting library built using Gaussian peak fitting on T₂ spectrum.
At last, several other kinds of edible oils, including oils already in the fingerprinting library and oils not in the fingerprinting library, are measured and the obtained FID curves are post processed using the mentioned methods for building the fingerprinting libraries. Oil discriminating is performed using Naive Bayes Classification with post processing.
The result shows that the FID fingerprinting library can distinguish different edible oils with the best resolution, compared to other three libraries. In conclusion, the presented method by simply building the FID fingerprinting library to identify different kinds of edible oils, can greatly improve the resolution of LF-NMR instrument on edible oils identification.