Prediction of acrylamide levels in baked bread using a portable near-infrared spectroscopy sensor

This study aims to determine acrylamide levels in different types of baked bread using a portable near-infrared (NIR) spectroscopy sensor. Two commonly consumed bread varieties, peasant-style loaves (n = 121) and wholemeal bread samples (n = 128), were analysed. Samples were scanned using a handheld NIR spectrometer (MicroNIR Onsite), and acrylamide content was quantified via liquid chromatography–mass spectrometry (LC-MS). For each bread type, two different principal component analysis (PCA) and partial least square discriminant analysis (PLS-DA) models were independently developed to classify samples based on their acrylamide level, using the benchmark level established by the European Commission (50 μg/kg) as a reference. PCA results revealed in both types of bread clustering of samples according to acrylamide content and spectral contributions around 980, 1200, and 1450 nm, bands associated with water, starch, protein and amino acids. The PLS-DA model for peasant-style loaves achieved 89 % accuracy in the validation, while the model for wholemeal bread reached 99 %. Sensitivity and specificity consistently exceeded 88 %, demonstrating that the method reliably distinguished samples below and above the EU benchmark level with high accuracy across all analyses. These findings demonstrate that a portable NIR sensor, combined with multivariate analysis, offers a promising rapid screening tool for non-destructive, on-site acrylamide detection in bakery products. Overall, this study provides the first evidence that portable NIR sensing can classify acrylamide levels in baked wheat bread in line with official EU benchmark thresholds, underscoring its potential for real-time industrial monitoring.