Scutellaria lateriflora (skullcap) is a medicinal herb that has a long history of use in the treatment of ailments such as insomnia and anxiety. Commercial herbal formulations claiming to contain S. laterifolia herba have flooded the consumer markets. However, due to intentional or unintentional adulteration, cases of hepatotoxicity have been reported. Possible adulteration with the potentially hepatotoxic Teucrium spp., T. canadense and T. chamaedrys has been reported. In this study, hyperspectral imaging in combination with multivariate image analysis methods was used to differentiate S. laterifolia, T. canadense, and T. chamaedrys raw materials in a non-destructive manner. Furthermore, the ability to detect adulteration of raw materials using the developed multivariate models was also investigated. Chemical images were captured using a shortwave infrared pushbroom imaging system in the wavelength range 920-2514 nm. Principal component analysis was applied to the images to investigate chemical differences between the species. Partial least squares discriminant analysis was used to model pre-assigned class images, and the classification model predicted the levels of adulteration in spiked raw materials. UHPLC-MS as an independent analytical technique was used to confirm chemical differences between the three species. The ability of hyperspectral chemical imaging as a non-destructive technique in the differentiation of the three species was achieved with three distinct clusters in the score scatter plot. A 92.3 % variation in modelled data using PC1 and PC2 was correlated to chemical differences between the three species. Near infrared signals in the regions 1924 nm and 2092 nm (positive P1), 1993 nm and 2186 nm (negative P1), 1918 nm, 2092 nm, and 2266 nm (positive P2), as well as 1993 nm and 2303 nm (negative P2) were identified as containing discriminating information using the loadings line plots. Chemical imaging of spiked samples showed spatial orientation of contaminants within the powdered samples, and percentage adulteration was accurately predicted at levels ≥ 40 % adulteration based on pixel abundance.