Fish freshness is regarded as one major parameter for seafood quality. However, it is lost inevitably in practice after catching and fish death, owing to the natural autolysis processwhich, in turn, trigger the growth of microorganisms and, consequently, the progressive loss offood characteristics and quality. This phenomenon is perceptible by changes in the sensorycharacteristics such as appearance, odour, taste and texture of fresh fish as well as in chemical, biochemical and microbiological changes.Fish market prices is highly depended on accurately predict its freshness and shelf-life.Predicted storage time in ice is defined as the number of days that the fish has been stored inice and it is possible to use these results to estimate the remaining shelf life. Quality IndexMethod (QIM) is currently the most wholesome and straightforward method of describingfreshness. However, it is time consuming and subjective and it is not always suitable for largescaleapplications.NIR spectroscopy has been proven to be a rapidly and non-destructive method forevaluating fish components (moisture, protein, fat, …) as well as it has shown good predictionserrors associated with fish storage time prediction. The purpose of this research is to test thepossibility of using NIR spectroscopy for non-destructively predicting freshness levels of plaicefish in Flanders, Belgium.In the preliminary study, spectroscopic measurements were performed on tested plaicesamples (n=10) subjected to different storage times assisted with Partial Least SquaresDiscriminant Analysis (PLSDA) indicated that NIR spectroscopy had great potential for nondestructiveplaice freshness discrimination. In the next step, the main study employing NIRdiffuse reflectance measurements for plaice samples (n=90) graded using commercial QIMscoring method at ILVO (Flanders, Belgium) together with Partial Least Squares (PLS) regression culminated on two different calibration models for predicting freshness expressed as storage days in ice (converted from the graded QIM scores): one for dark skin measurements with prediction performances of 1.82, 2.22 and 0.804 for RMSECV, RMSEP and R2p, respectively, using the selected wavelength range of 1400 to 1580 nm; and one for white skin measurements with those parameters of 2.356, 2.59 and 0.677 for RMSECV, RMSEP and R2p,respectively, using the full wavelength range studied of 940 to 1700 nm.
|Date of Award||28 Dec 2017|
- Universidade Católica Portuguesa
|Supervisor||Wouter Saeys (Supervisor) & António Silva Ferreira (Co-Supervisor)|
- NIR spectroscopy
- Mestrado em Engenharia Alimentar