Portable and handheld hyperspectral imaging for non-destructive acrylamide prediction in ground potato crisps

The most common analytical techniques to determine acrylamide in foods are costly, time consuming and destructive, therefore not suitable for process control during food manufacturing. Hyperspectral imaging (HSI) presents a promising non-destructive alternative for real-time acrylamide detection. This study investigated the feasibility of two portable HSI systems, a carry-on portable hyperspectral camera and an all-in-one spectral imaging (ASI) setup capturing images in 400-1000 nm range, for acrylamide prediction in potato crisps. Prediction models developed using HSI data were compared with models obtained by high-end Fourier-transform infrared (FTIR) spectroscopy instrument and showed better prediction performance. The HSI systems were then evaluated across a larger sample set and further tested on an independent sample set of commercial crisps with low acrylamide content. Partial least square regression (PLSR) model calibrated within medium range of acrylamide levels (< 3000 μg/kg) achieved root mean square error of calibration (RMSEC) of 314.5 μg/kg (r = 0.9) and root mean square error of prediction (RMSEP) of 466 μg/kg (r = 0.2). 6 key wavelength bands were identified and used to build a model with RMSEC of 267.9 μg/kg (r = 0.9) and RMSEP of 312.7 μg/kg (r = 0.9). These results indicate that portable and handheld HSI systems can be used as rapid non-destructive technique to predict acrylamide in potato crisps.