Deep Learning-Assisted Prediction of Acrylamide Concentration In Potato Crisps with Electrochemical Impedance Spectroscopy Data of Raw Tuber

Acrylamide (AA) is known by its cytotoxic, carcinogenic, and mutagenic effects. AA is present in thermal processed food, such as potato crisps, where AA contents can be high. Thus, the analysis of the quality of raw materials before the thermal process, with low content of acrylamide precursors such as asparagine and reducing sugars, can help manufacturers to take decisions on the acceptance of the materials. In this sense, this work proposes the analysis of raw potato by electrochemical impedance spectroscopy to predict the acrylamide concentration in potato crisps using a deep learning model. With this model, an acrylamide concentration prediction in potato crisps in a range of 300–700 μg kg⁻¹ was achieved, below Benchmark recommended level of 750 μg kg⁻¹ from European Commission in potato crisps. This fast, simple, and low cost system does not require sample treatment, facilitating its application in raw material quality control.