Support Vector Regression for quantitative determination of human chorionic gonadotropin concentration from gold immunochromatographic strip

There are several methods for quantitative determination of human chorionic gonadotropin (hCG), among these methods, the gold immunochromatographic assay has the advantages of simple operation, low costs and short operation time. However, this assay can only get qualitative or semi-quantitative results when observed directly with naked eyes. By combining the fuzzy C-means algorithm and the Support Vector Regression (SVR), this paper presents a rapid quantitative determination method of hCG immunochromatographic assay strip. In this paper, SVR is used to predict the hCG concentration. In the experiment, the data derived from 35 strips CCD image are used for SVR model training; on the other hand, the data derived from other 18 strips are used for model testing. The results show that the SVR yields a good result, the total MSE of the 18 strips is 10.2. The CV is 7.07%. This method is proved to be practical and objective, as well as enhances the detection sensitivity in some extent.