Rapid Identification Of Pseudomonas In Chicken Based On Olfaction Visualization And Near Infrared Spectroscopy

Pseudomonas spp.is the main bacteria involved in chicken degradation,which ultimately affects the chicken quality and potentially poses public health threats.The conventional detection methods involve complex preprocessing steps;time,labour and reagent demanding;have low recognition accuracies,with strains detection constraints,which are often unsuitable for rapid real-time detection.This study utilized the olfaction visualization technology,near infrared(NIR)spectroscopy,and the combination of the two systems to attempt the rapid real-time detection of Pseudomonas spp.in degrading chicken.The main focuses of the study are summarized as follows:1.Isolation and identification of the dominant spoilage bacteria in degrading chicken.First,18 strains of bacteria were isolated from degrading chicken with different selective culture-medium.The isolated bacteria were then identified via Polymerase Chain Reaction(PCR)technology,from which four strains of Pseudomonas spp.(i.e.,Pseudomonas gessardii,Pseudomonas psychrophila,Pseudomonas fragi and Pseudomonas fluorescens),the main bacteria involved in chicken degradation,were selected as the research object for the real-time identification experiment.2.Study on rapid identification of Pseudomonas spp.in chicken by olfaction visualization technology.Initially,12 different kinds of gas sensors made up of 9 porphyrin dyes and 3 pH indicators sensitive to odors from Pseudomonas spp.liquid were selected to fabricate the olfaction visual sensor.A color change profile for each sample was obtained by computing the differentials in the images of sensors array prior and post exposure to the head-gas of prepared liquid samples containing the four isolated Pseudomonas spp.and their combined mixture samples.The values of RGB(i.e.,red,green and blue)color components were extracted from each dye in color change profiles,thus 36 color variables of each sample were obtained.Principal component analysis(PCA)was performed on the original 36 variables and then chemometric alogorithms such K nearest neighbor(KNN),support vector machine(SVM)and back propagation artificial neural network(BP-ANN)were used to build identification models for the different samples of Pseudomonas spp.The recognition rate of the BP-ANN was found to be superior to the KNN and SVM model,with 92.5% and 90% obtained in its training and prediction set respectively using Principal Components(PCs)number 6.3.Study on rapid identification of Pseudomonas spp.in chicken by NIR spectroscopy.The information of the transmission spectrum of five groups Pseudomonas spp.liquid samples were collected via NIR spectroscopy.The original spectrums were preprocessed by standard normal variable transformation(SNV),and then analyzed by PCA.After which 10 times genetic algorithm(GA)were operated to select spectral characteristic variables to build KNN,SVM and BP-ANN identification models for the different samples of Pseudomonas spp.The recognition rate of the BP-ANN was superior to the KNN and SVM model,with 98.33% and 95% recognition rates recorded in the training and prediction set respectively,where the number of spectral characteristic variables and PCs were set to 33 and 8 respectively.4.Study on rapid identification of Pseudomonas spp.in chicken by multi-sensor information fusion of olfaction visualization technology and near infrared spectroscopy.Based on the successfully used of single technique in identification of Pseudomonas spp.in chicken.The study attempted the fusion of data obtained from the two technologies for the comprehensive detection of five groups of Pseudomonas spp.A total of 36 color and 33 spectral characteristic variables were extracted from the data collected by olfaction visualization technology and NIR spectroscopy simultaneously.The number of PCs of the color and spectral characteristic variables were optimized by PCA selected as the input variables for BP-ANN model,with its best performance of achieved with the PCs number of color characteristic variables and spectral characteristic variables at 6 and 8 respectively.A high recognition rate of 100% in training set and 98.75% in prediction set were achieved.This demonstrated that the fusion model performance improved significantly over the single model indicating higher accuracy and stability.The study provides theoretical foundation and technological guidance for the application of multi-information fusion technology for the detection of Pseudomonas spp.in chicken.It thus consequently holds practical significance for the safeguarding chicken quality.