| Sumario: | A simple, rapid procedure is required for the routine detection and quantification of haemolysis, one of the main sources of unreliable results in serum analysis. In this study, we compared two different approaches for the rapid determination of haemolysis in cattle serum. The first consisted of estimating haemolysis via a simple direct ultraviolet–visible (UV–VIS) spectrophotometric measurement of serum samples. The second involved analysis of red, green, blue (RGB) colour data extracted from digital images of serum samples and relating the haemoglobin (Hb) content by means of both univariate (R, G, B and intensity separately) and multivariate calibrations (R, G, B and intensity jointly) using partial least squares regression and artificial neural networks. The direct UV–VIS analysis and RGB-multivariate analysis using neural network methods were both appropriate for evaluating haemolysis in serum cattle samples. The procedures displayed good accuracy (mean recoveries of 100.7 and 102.1%, respectively), adequate precision (with coefficients of variation from 0.21 to 2.68%), limit of detection (0.14 and 0.21 g L–1, respectively), and linearity of up to 10 g L–
|
|---|
