Study On Text Mining Algorithm For Ultrasound Examination Of Chronic Liver Diseases Based On Spectral Clustering

Ultrasound examination is to irradiate the lesion with a weak ultrasonic instrument,generate the image of the echo of the lesion tissue and process it,and the doctor reads the ultrasonic examination image formation report text.Ultrasound images are less likely to be disturbed by respiration,so the diagnostic accuracy of ultrasonography for chronic liver disease is high.In recent years,the number of patients with chronic liver disease in China has been increasing,and a large number of ultrasound examination report texts have been accumulated.The paper adopts an improved spectral clustering algorithm to cluster the ultrasound examination report texts with high-dimensional data features,and mines the potential value of the ultrasound examination report texts to provide technical support for the prediction and diagnosis of chronic liver diseases.The main work is as follows:1.This paper analyzes the characteristics of spectral clustering algorithm which is difficult to determine the number of clusters,and thus the clustering efficiency is low.The two clustering numbers K values are obtained by using the elbow method and the contour coefficient method respectively,which are used to determine the number of clusters.Value interval.The paper selects 12 types of 1200 actual text data on the Sogou News Chinese corpus,and verifies that the actual number of cluste rs is within the value interval calculated by the paper.The paper compares the clustering effect of spectral clustering and k-means algorithm.The results show that the Rand index obtained by the spectral clustering algorithm is 11.34% higher than the k-means algorithm,and the Jaccard index is increased by 14.67% compared with the k-means algorithm.2.The paper selected 110 ultrasound examination report texts of patients with real chronic liver disease.After text preprocessing,the above algorithm was u sed.The selected Calinsky criterion is the effectiveness index of the cluster and the optimal cluster number K is determined to be 5.The paper produced high-frequency word cloud maps,and analyzed the relationship between high-frequency words and diagnostic results of the five types of ultrasound reports,and provided support for improving the accuracy of ultrasound diagnosis of chronic liver diseases.In this paper,the spectral clustering algorithm is partially improved,and a method for accurately defining the number of clusters is proposed.Through the actual application of chronic liver disease ultrasound report samples,the validity of the improved algorithm of the paper is verified,and it has certain practical application value.