Identification Of Pathogens In Bloodstream Infection Based On Adaptive Clustering Algorithm

Bloodstream infection is a serious infectious disease in the world due to its high morbidity and mortality.Once being suspected as clinical bloodstream infection,the first time to understand the types of pathogens is conducive for clinicians to choose antibiotics accurately and rationally.Blood culture was a conventional method to identify pathogens,but it requires substantial amount of time and prevents appropriate treatments from being administered immediately.In this paper,NIR spectroscopy combined with Multiclass Relevance Vector Machine(M-RVM)was proposed to identify pathogens.In addition,an adaptive clustering algorithm was used to verify the unknown samples,which is mistakenly identified by M-RVM.Finally,the effect of characteristic wavelength optimization on the discriminant model was discussed.This paper mainly completed the following research:Firstly,the characteristics and harm of bloodstream infection were introduced.The defects of the existing microbiological methods were pointed out,and then the identification of the pathogens by vibration spectroscopy was presented.At the same time,the qualitative analysis technology of infrared spectroscopy and the status quo of spectroscopy combined with pattern recognition in microbial identification was described.Secondly,the principles and advantages of M-RVM algorithm were expounded.The pathogens of Escherichia coli,Staphylococcus aureus and Pseudomonas aeruginosa were collected by near-infrared spectroscopy.Then,the corresponding classifiers were constructed by M-RVM and the accuracy of the classifier for different kernel functions is analyzed.Moreover,the precision of M-RVM,PLS-DA and SVM were compared under different spectral pretreatments.In order to solve the problem of correct identification of unknown samples,an adaptive clustering algorithm was used to further analyze the discriminant results.Whether the new samples belong to the modeling category was according to the number of clustering centers.The clustering center process is automatically determined by regression analysis.Additionally,in order to obtain the optimal clustering result,the input parameter cutoff distance of the algorithm was optimized by using particle swarm optimization algorithm(PSO).Finally,the identification of Pseudomonas aeruginosa by established M-RVM classifiers were further validated by adaptive clustering algorithm.The results showed that Pseudomonas aeruginosa was different from the two model samples,which demonstrated the validity of the algorithm.Finally,due to the large number of spectral wavelengths and too many useless information and noise were contained,the extraction of effective characteristic wavelengths will simplify the identification model and improve the accuracy of the model.In this paper,swarm intelligent optimization algorithm combined with partial least squares were used to select wavelength variables.The partial least squares and M-RVM discriminant models were constructed by using the characteristic wavelengths,and the accuracy of the model was analyzed and compared before and after the selection of the wavelength variables.