Short-wave infrared hyperspectral imaging coupled with ECOC-SVM classification enables discrimination of hemp plants treated with PFOA and PFOS

This study investigates the potential of Short-Wave Infrared Hyperspectral Imaging (SWIR-HSI: 1000–2500 nm) combined with chemometrics to discriminate organ-specific spectral responses in Cannabis sativa plants exposed to perfluorooctanoic acid (PFOA) and perfluorooctane sulfonate (PFOS), two of the most widespread per- and polyfluoroalkyl substances (PFAS) in the environment. Leaves, flowers, stems, and roots of plants grown under controlled hydroponic conditions were analyzed. Fourier-transform infrared (FT-IR) spectroscopy was used to support the interpretation of spectral variations in plant tissues following PFAS exposure. FT-IR analysis revealed treatment-dependent changes in major biochemical components, such as lipids, proteins, and polysaccharides, associated with variations in key absorption regions related to O–H, C–H, and C=O functional groups. SWIR-HSI data, supported by FT-IR analysis, revealed organ-dependent spectral responses to PFAS exposure across the different plant tissues analyzed, and multivariate exploratory analysis (PCA, HCA and t-SNE) highlighted varying degrees of separation between control and treated samples. Based on these observations, binary ECOC-SVM models were developed to assess the reproducibility of the spectral patterns and to discriminate between control and PFAS-treated samples. The models achieved high classification performance, with sensitivity and specificity up to 0.95–1.00 in calibration and cross-validation, and 0.79–0.87 in independent prediction, with slightly lower performance observed for PFOS-treated samples. Overall, these results suggest that the proposed SWIR-HSI-based strategy represents a preliminary approach for the non-destructive detection of PFAS-associated spectral variations in plant organs, providing a basis for further investigation of PFAS-induced plant responses.