Algorithm Optimization Of Raman Spectroscopy Screening For Cervical Diseases Based On Deep Belief Network

Cervical diseases are one of the common diseases affecting female reproductive health,and there is an urgent need to raise awareness and prevention measures.Therefore,in this paper,Raman spectral data of cervical diseases collected by the laboratory are selected and a variety of artificial intelligence algorithms are used for three-classification analysis,expecting to provide scientific basis for the early diagnosis and treatment of cervical diseases,as well as diagnostic reference for medical workers.This paper is divided into the following parts:(1)Pretreatment and analysis of Raman spectral dataThis part introduces the pre-processing and analysis of Raman spectral data.Raman spectroscopy is a non-damaging spectral technique that can provide information on the vibration of chemical bonds in the sample,thus supporting the classification of cervical diseases.In the process of spectral data acquisition,electronic noise,background stray light and other factors are inevitably introduced,so it is very important to preprocess the data.In the process of preprocessing,SG smoothing is used to reduce the noise of spectral data.Then baseline correction is performed to reduce the fluorescent background,and finally normalization is carried out,so that the original data becomes the pre-processed data that can be classified by the classifier.(2)Comparison between SVM multi-classification model and other algorithmsAfter the preprocessing,this paper uses the support vector machine(SVM)algorithm to build the Raman spectrum classification model of cervical pathological diagnosis.Since uterine fibroids are the early stage of cervical cancer,the classification of uterine fibroids is introduced to try to screen cervical cancer early.SVM is a classification algorithm based on statistical learning theory,which is widely used in image recognition,biological information processing and other fields.It has generalization ability and good classification accuracy.Compared with linear discriminant analysis classification model and K nearest neighbor classification model,the results show that SVM classification results are far superior to the other two models,and also prove the possibility of early diagnosis of cervical cancer.(3)Optimization of DBN algorithm and sparrow search algorithmIn addition to the SVM multi-classification model,a special machine learning method,deep confidence algorithm(DBN),was also used in this paper to conduct three-classification analysis of cervical disease samples.DBN is a deep learning algorithm,which uses the greedy layer-by-layer training method to build the network layer by layer.By optimizing the network parameters,the ability to reconstruct and classify the input data is obtained.Compared with SVM algorithm,DBN algorithm has better performance in processing large amount and high dimensional data.In the experiment of this paper,DBN algorithm can not only improve the classification accuracy of cervical disease samples,but also has a nearly 12.13%higher accuracy than SVM algorithm.Of course,in order to further optimize the performance of DBN algorithm,this paper also tries to optimize the sparrow search algorithm to get better classification effect and prediction accuracy.After optimization,the classification accuracy and operation efficiency of the algorithm have been improved to some extent.Due to the small number of data samples,10 dB,20 dB and 30 dB Gaussian noise were added on the basis of the original samples to further illustrate the significance of the results.The results showed that the classification results were not affected by Gaussian noise.The results show that,compared with SVM algorithm,DBN algorithm has better performance in the multi-classification diagnosis of cervical diseases,no matter in the running time of algorithm,classification accuracy of algorithm and other indicators.