Computer-aided technologies for food risk assessment

Nowadays, the relationship among food, health, and economy is an emerging topic that engages the modern global society from an interdisciplinary perspective. The Food Risk assessment has been formalized and incorporated into the specific discipline addressed to everyone involved with food from production to consumption, including growers, processors, regulators, distributors, retailers and consumers. However, both intentional and unintentional actions committed for economic gains could make an attempt on people’s health. In recent years, many tools have been developed to help the authorities involved in controls and consumers too. The integration of multidisciplinary techniques has favorably supported the study and the development of tools related to the field of the Systems Biology as well as the application of state-of-the-art techniques deriving from other application fields such as the Computer Science. To counteract and operate with reaction and prevention in my Ph.D. I investigate the use of original Computer-Aided technologies in two particular instances. The first one refers to a Food Traceability issue related to dairy product control. I studied and implemented a heuristic procedure that allows food inspectors to highlight possible adulterations in cheese production into the small farm environment. The procedure is mainly based on Short Tandem Repeat investigation to compare the DNA fingerprint among cows, milk, and cheese. The second one regards the Food Fraud discipline. I developed a mobile application to counteract the problem of fish species substitution and mislabelling. The infrastructure implemented is composed of a cloud remote server where both image analysis and machine learning algorithm take part. The main breakthrough on this topic has been reached with a deep learning classification system which allowed to obtain an improvement in the global accuracy to correctly identify the fish species. Eventually, in the last topic, I deal with the problem of fish fillets identification. The main outcome of this preliminary study is the application of a portable Near Infra-Res molecular sensor that was specifically trained to discriminate the fish fillets available in a sample database.