Near-infrared hyperspectral imaging as a rapid tool for detection of toxin-producing microalgae in seawater

Near-infrared hyperspectral imaging (NIR-HSI) was evaluated as a rapid, non-destructive screening tool for the detection of harmful toxin-producing microalgae directly in natural seawater samples. One year of weekly to bi-weekly collected samples (May 2020-June 2021) from two stations in Ria de Aveiro lagoon (Portugal) were analysed, covering strong seasonal and inter-annual variability in phytoplankton assemblages. Hyperspectral images of filtered seawater samples were recorded in the 900-1700 nm range and processed by multivariate image analysis. Cell counts and taxonomic identification were obtained using the reference Utermöhl inverted microscopy method for species producers of AST, PST, DST, YTX and AZT toxins. PLS-DA models built on mean image spectra enabled to distinguish between samples above and below the detection limits of the reference method for all five toxin-producers. Cross-validated accuracies reached 0.98 for AST and PST producers and 0.87-0.90 for DST, AZT and YTX producers. NIR-HSI demonstrated superior performance compared with FT-MIR spectroscopy for AST, PST and DST groups. Furthermore, NIR-HSI allowed species-level discrimination within Dinophysis, with accuracies up to 0.98 for D. ovum. The results demonstrate that NIR-HSI can detect toxic microalgae at ecologically relevant abundances and has strong potential as a screening technology for HAB monitoring programmes.