Robust detection methodology of milk heat treatment in cheese based on volatile profile fingerprinting

Abstract

The aim of the study was to develop an approach to discriminate whether cheese is produced from raw or heat-treated milk. The method was based on multivariate discrimination of volatile organic compounds in the cheeses' headspace. Although the method was developed to detect issues with food authenticity for a Dutch traditional speciality guaranteed-protected cheese, this principle is likely to be applicable to detect heat-treatment in milk in other raw-milk cheese types. The multivariate classification method was aimed to be robust, employing an ensemble classification, based on six independent classification algorithms that require little or no tuning. The method was validated using a recently developed validation protocol designed for multivariate classification methods with a large validation set that was gathered separately from the training set. Based on the method's “worst-case”-classification performance, an overall 88% correct classification is expected.