A chemometric approach to assess the oil composition and content of

Oil and fatty acid content

The average oil content measured using AOAC²⁸ for untreated mustard seeds was 26.41 (± 0.18) % and MW treated seeds of MW1 and MW2 treatments were 33.71 (± 0.98) and 32.28 (± 0.12) %, respectively; comparable values have been reported for rapeseed oil in the past¹². The fatty acid composition of (Table 1) MW treated samples as compared with the untreated samples demonstrated an increase of fatty acids like erucic acid (C22:1), linoleic acid (C18:2) and overall monounsaturated fatty acids (MUFA); while a reduction was observed in other fatty acids like alpha-linolenic acid (C18:3), eicosenoic acid (C20:1), oleic acid and polyunsaturated fatty acids (PUFA). Similar observations have been made by other researchers as well^17,36.

The proportional composition of fatty acids like unsaturated and PUFA in MW treated and untreated remained unaffected significantly (p 

Spectral profile

The average reflectance spectra (Fig. 1a,b) obtained for mustard seeds using Vis–NIR–SWIR HSI showed different shapes and reflectance intensity in line with the experimental spectral range. The effect of microwave treatment probably affected the seed reflectance intensity due to reduced enzyme activity and denaturation of some proteins in the mustard seeds³⁷. It was observed that the variations in reflectance values among treatments were greater in the Vis–NIR spectral region than in the SWIR region. In Vis–NIR spectra with SG smoothing (Fig. 1a) the curves depicted a steep drop from 400 to 441 nm, it was observed that the variations in reflectance values among treatments were greater in the Vis–NIR spectral region than in the SWIR region. In Vis–NIR spectra with SG smoothing (Fig. 1a) the curves depicted a steep drop from 400 to 441 nm, followed by a gradual lowering up to 557 nm, then a steady increase was observed up to 1000 nm, which was predominantly caused by the third overtone of C–H stretching³⁸ and relatively flat profile was observed with SNV and SG smoothing (Fig. 1c). In the case of the SWIR hyperspectral region SG smoothing (Fig. 1b) and SNV with SG smoothing (Fig. 1d) revealed prominent reflectance bands at 1140, 1164, 1374 and 1597 nm. The spectral range of about 1164 and 1374 nm³⁹ relates to the C–H stretching vibration elongation in the second overtone (–CH₂), which can be ascribed to the oil content⁴⁰. Also, the O–H stretching in the first overtone is typically associated with water and its peak is observed at 1597 nm, in this case, it is likely that the peak at 1597 nm is related to cellulosic components instead of water due to the low moisture level of the samples (about 7%). Therefore, even though the peak is typically associated with water, in this specific case, it is more likely connected to cellulosic components. The performance of classification of the prediction models was bolstered by the presence of these peaks in the SWIR region, due to this the performance of the Vis–NIR HSI spectrum was lower than that of the SWIR HSI spectrum.

Mean spectra of mustard seeds after smoothing with SG (a,b), SNV and SG smoothing (c,d) and SG-D2 (e,f) to Vis–NIR and SWIR spectral data, respectively.

After performing an SG-D2 spectral pre-processing (Fig. 1e,f), additional peaks were observed throughout the spectral data. This phenomenon can be attributed to the inherent inverse relation between the amplitude and width of the successive derivatives of a wave-form; in this case it was the second derivative⁴¹. In Vis–NIR spectra which primarily represented the colour differences of the samples, the peaks and valleys were observed at 411 and 429 nm (violet band), 484 nm (blue band), 613 nm (yellow band) and 951, 990 nm⁴². The changes in the peaks and valleys can be attributed to the MW treatment of the mustard seeds, which may have caused alterations in their colour. In SWIR spectra, additional peaks and valleys were observed at 1160 and 1208 nm, representing the second overtone C–H. The observation of additional peaks and valleys in the C–H overtone region in the SWIR spectra suggests that MW treatment may have caused alterations in the chemical structure of the mustard seeds. These changes could be attributed to the heating effects of MW treatment, which may have caused the breakdown or rearrangement of chemical bonds⁴³.

Principal component analysis (PCA)

The Vis–NIR–SWIR hyperspectral data was pre-processed and subjected to principal component analysis to depict variance across samples and to assess the influence of spectral pre-processing on the classification of MW-treated and untreated mustard seeds. The spectral data matrices for PCA contain 2700…