Spoilage of fresh fish products by the action of bacteria is one of the main causes of the short shelf-life of these products. A range of bacteria are responsible for this and are referred to collectively as "spoilage bacteria". Currently methods to detect both spoilage of the product and the presence of number of bacteria are time-consuming, for example requiring 24-hour incubation periods, or require specialised labour such as tasting panels. Near infra-red spectroscopy (NIR) is widely used in the food industry to monitor factors such as fat and moisture content in a range of foods. Although it has been used to distinguish different types of bacteria and, in a few cases, to quantify the number of bacteria in different materials, there is a lack of information on the ability of the method to quantify bacteria directly on food products. The aim of this project was to determine whether NIR had the potential to be used as a method to detect and predict microbial spoilage of fresh fish products.

NIR was easily able to distinguish between fresh Atlantic Salmon fillets and those stored for nine days at 4°C indicating that NIR can detect spoilage. Partial least squares regression prediction models for the number of total bacteria and the number of Enterobacteriaceae present were developed. These models used the NIR spectra collected when the fish was fresh to predict the number of bacteria that would be present nine days later. There are many factors (protein and fat content of the salmon itself for example) that contribute to the differences in the NIR spectra that are unrelated to the numbers of bacteria. Hence for any model to be useful it needs to include as many of these variables as possible. In conclusion, the results of this project show that NIR has potential to be a useful method for detecting and predicting bacterial levels on fish but much more work is required to develop a suitably robust model.