Rapid Detection Of Three Antibiotics In Goat Milk By Electronic Nose

Milk is an almost nearly complete natural food, which has became one of the most important sources of nourishment for people. Goat milk as a burgeoning milk sources in China, which has high nutrition value and health function, has gained increasing attention. But with the increasing demand of the milk and its products, in dairy cattle management, antibiotic drugs are widely used for ensure production output as well as preventing and curing disease. The application of veterinary, such as penicillin, with improper administration can result in antibiotic residues in milk and milk products, which are the major constraints to dairy industry. As a potential threats to public health, antibiotic residues present in milk can induce patients with hypersensitivity to allergic shock even death. Presently the major testing techniques of the antibiotic residues are instrumental analysis, such as GC, LC, HPLC and GC/LC-MS, and microbiological methods. These techniques could monitor trace residues with high sensitivity and accuracy, but there are also some downsides, such as complicated operation, low speed and high cost, which could not sufficient to meet the pressing needs in daily production or life.In this paper,A PEN3electronic nose was employed to collected the volatiles emitted from the goat milk with4concentrations (0,50,100,200μL/L) antibiotics(penicillin, gentamycin, streptomycin) at the same time. Loadings, LDA were applied to analyze the result and the responses signals were chose for the pattern recognitions by fisher discriminatory analysis and multilayer perceptron neural network. The main conclusion is as follows:The diagrams of the response signals showed the volatiles from the goat milk with penicillin and streptomycin were restrained as well as gentamycin boosted the activity of volatiles. The results of Loadings showed the sensors W1W and W2W, which are sensitive to sulfide, played a central role in evaluating the concentrations of the antibiotics. So sulfide is an important influencing factor in classifying.The results of LDA showed that the response signals of the goat milk was prone to distinguish the concentrations of antibiotics. And it could clearly discriminate the control group from the others. Although the classifications had some overlap, the LDA has a good effect of the distinction of the concentrations.Fisher discriminant analysis and multilayer perceptron neural network were applied to predict the concentrations of penicillin G, gentamycin sulfate, streptomycin sulfate in goat milk. Generally, the correct rates of the training group and the testing group of FDA were both more than98%, and the MLPNs’were both more than94.9%. The results showed that forecast precision to the concentrations of three antibiotics in goat milk from two kinds of models were both very high. Therefore fisher discriminant analysis and multilayer perceptron neural network can be used as effective forecasting tools for using the signals of electronic nose to predict the concentrations of three kinds of antibiotics in goat milk.