Agricultural crops of high value are frequently targeted by economic adulteration across the world. Saffron powder, being one of the most expensive spices and colorants on the market, is particularly vulnerable to adulteration with extraneous plant materials or synthetic colorants. However, the current international standard method has several drawbacks, such as being vulnerable to yellow artificial colorant adulteration and requiring tedious laboratory measuring procedures. To address these challenges, we previously developed a portable and versatile method for determining saffron quality using a thin-layer chromatography technique coupled with Raman spectroscopy (TLC-Raman). In this study, our aim was to improve the accuracy of the classification and quantification of adulterants in saffron by utilizing mid-level data fusion of TLC imaging and Raman spectral data. In summary, the featured imaging data and featured Raman data were concatenated into one data matrix. The classification and quantification results of saffron adulterants were compared between the fused data and the analysis based on each individual dataset. The best classification result was obtained from the partial least squares-discriminant analysis (PLS-DA) model developed using the mid-level fusion dataset, which accurately determined saffron with artificial adulterants (red 40 or yellow 5 at 2-10%, w/w) and natural plant adulterants (safflower and turmeric at 20-100%, w/w) with an overall accuracy of 99.52% and 99.20% in the training and validation group, respectively. Regarding quantification analysis, the PLS models built with the fused data block demonstrated improved quantification performance in terms of R2 and root-mean-square errors for most of the PLS models. In conclusion, the present study highlighted the significant potential of fusing TLC imaging data and Raman spectral data to improve saffron classification and quantification accuracy via the mid-level data fusion, which will facilitate rapid and accurate decision-making on site.
Read full abstract