Research On Intelligent Processing Algorithm For Digital Pathological Image Of Adenocarcinoma

In pathological analysis,digital pathological images are used as an important reference for clinical diagnosis and treatment because they contain a large amount of rich pathological information.Therefore,the research of intelligent processing algorithms for digital pathological images is a very important and necessary task,and one of the hot spots in the field of medical research.The current research direction is mainly from computer-aided diagnosis,through pre-processing,image segmentation,feature extraction,classification to obtain pathological diagnosis information to help experts analyze pathological images.In recent years,lung cancer has become the cancer with the highest morbidity,highest mortality,and fastest growth rate.Adenocarcinoma is a type of lung cancer,and its incidence accounts for 50% of the total lung cancer.Therefore,this paper takes adenocarcinoma digital pathological image as the research object to analyze its intelligent processing algorithm.This study starts with extracting features of digital pathological images,and studies the application of the extracted features in classification detection and genetic analysis of pathological images.The main contents of the research are: first,the HOG algorithm is improved.Then use the improved algorithm to obtain the data information included in the pathological image.Subsequently,the extracted data information is used to train support vector machines(SVM)to complete the classification detection of adenocarcinoma pathological images.Finally,the weighted gene co-expression network(WGCNA)method is used to study the molecular layer information of pathological images.The key information contained in it is the target gene,which has great biological significance for the pathological research of adenocarcinoma.The main content of this article is as follows:(1)In the feature extraction step of adenocarcinoma pathological images,based on the calculation principle of the original HOG feature operator,two direction vector matrices are introduced.In this way,the feature extraction directions of the original algorithm,that is,the horizontal direction and the vertical direction are expanded into horizontal,vertical,and two diagonal directions to realize the extraction of feature information.This improvement greatlyimproves the entropy value of feature information,and fully extracts the pathological information contained in the pathological image.(2)Research on image classification algorithm based on the obtained feature information.The particle swarm optimization algorithm(PSO)is applied to the SVM classifier to optimize the two parameters c and g of the SVM,so as to achieve the purpose of optimizing the SVM and improve the accuracy of the classifier's detection of image types.In order to make the PSO optimization algorithm get better optimization performance,the newly proposed PSO algorithm redesigns the position update formula based on the original PSO optimization algorithm.In the position update formula,a new function and position weight w that increase the correlation between particles are introduced according to the principle of adaptability,and the speed update formula is discarded.This design improves the optimization performance of the PSO optimization algorithm and obtains good results when optimizing the SVM.(3)On the basis of obtaining the feature information and classification results of pathological images,in order to further understand the information on the molecular layer of adenocarcinoma pathology,this paper combines the WGCNA method with digital pathological images for analysis.The feature information and classification results of digital pathological images are used to estimate key genes related to adenocarcinoma diseases,providing more reliable diagnosis and treatment information for medical research.The analysis and improvement of the above algorithm and experimental verification show that the method proposed in this paper has certain effectiveness in feature extraction and classification of digital pathological images.In the case that the information entropy of the pathological image feature information reaches nearly 3.9,the classifier obtains 98.5%classification accuracy.Finally,the method of predicting key genes by combining WGCNA with digital pathological images was explored.