Development of analytical methods based on hyperspectral images and infrared spectroscopy to study the distribution and quantify active principles in pellets for salmon
Desarrollo de métodos analíticos basados en imágenes hiperespectrales y espectroscopía infrarroja para estudiar la distribución y cuantificar principios activos en pellets para salmones
Fecha
2019Autor
Castillo Felices, Rosario del Pilar
UNIVERSIDAD DE CONCEPCION
Institución
Resumen
The Chilean salmon industry is one of the largest worldwide. One of the biggest challenges in salmon farms is to effectively administer antibiotic therapy to salmon once there is an outbreak of any disease. One of the most used to ways to give antibiotics to the salmon is through salmon food, submerging food pellets already created with a mixture of oil with the antibiotic, a technique known as surface coating. With surface coating it is assumed that the antibiotic penetrates sufficiently to be distributed in a homogeneous way into the pellets. In addition, to confirm the dose of the antibiotic in the pellets, it is quantified by high-resolution thin layer chromatography (HPTLC), however, this analytical technique is expensive and very slow. Given the need to use analytical methods to study the distribution of an antibiotic in salmon feed pellets and quantify it more quickly and cheaply, it was proposed in this thesis that analytical methods based on hyperspectral images and infrared spectroscopy manage to study the distribution and quantify the Florfenicol antibiotic (FF) in pellets for salmon. The first result was the development and validation of multivariate methods with multivariate curve resolution (MCR-ALS), discriminant analysis with partial least squares (PLS-DA) and multivariate calibration of images with partial minimum squares (PLS), to analyze hyperspectral images obtained with infrared spectroscopy of pellets with FF created by Cargill©, resulting in FF distribution maps in the analyzed images. The analysis of these images, which were obtained from the surface towards the center of the pellet, indicates that only the images obtained near the surface of the pellets contain FF, and their distribution in this area is heterogeneous, so surface coating - coating is not a satisfactory technique to include FF to the pellets. The second result was the development and validation of a multivariate calibration model with PLS, with which it was possible to quantify FF in pellets for salmon created by Cargill © satisfactorily, with a prediction error of 1.74 mg FF/g pellet, in spite of FF is less than 1% of the pellets and they have highly active components in the infrared spectrum, such as carbohydrates, oils and proteins. Finally, we conclude that it was possible to give analytical solutions to two analytical needs of the salmon industry, which could be used to satisfy the needs of other entities that require the use of these methods. The high complexity of the pellets for salmon in the infrared spectrum was not an impediment to the development of the proposed methods due to the proposed multivariate analysis, in spite of, as was mentioned above, FF was found in low quantities within the pellets.