Fish species identification is normally undertaken using DNA barcoding methods. However, it is often useful to have an on-site rapid non-destructive method to verify fish species. In this study, 2 Raman spectrometers (a portable Raman spectrometer and a benchtop confocal Raman spectrometer) were used and compared for their performance to identify 11 species of fish (4 species of Salmonidae and 7 species of non-Salmonidae). Supervised chemometric/machine learning classification models were constructed based on a hierarchical classification principle to develop this 11-class identification. Both Raman spectrometers were able to differentiate Salmonidae from non-Salmonidae fish with close to 100% accuracy. To further identify the fish to species level, the portable Raman spectrometer provided better accuracy (i.e., 93% and 93% accuracy for the Salmonidae group and non-Salmonidae identification, respectively) compared to the benchtop Raman spectrometer (i.e., 90% and 84% accuracy for the Salmonidae group and non-Salmonidae identification, respectively). The overall non-destructive analytical time was only 5 minutes.
Read the full open access paper here