The Diagnosis Classification Of Dommon Pediatric Posterior Fossa Tumors Using Texture Analysis

Tumors are one of the major diseases that threaten human health.Among them,brain tumors are the most common solid tumors,with half of the posterior fossa occurring.The younger age of onset of tumors is even more important to human beings.Ependymomas,medulloblastomas,and astrocytomas are common pediatric posterior fossa tumors.For different tumors,the treatment and prognosis options are also different,so it is very important to correctly identify the tumor type before surgery.At present,doctors mainly identify MR images of tumors by artificial.Manual identification has a large workload,,has subjective characteristics and poor repeatability,and has certain limitations.There are large errors in empirical judgments based on human experience.In recent years,texture analysis has become the focus of computer-aided detection.The MR image contains rich texture information.As a non-invasive diagnosis,computer-aided detection relies on the computer to perform texture analysis on the MR images and extract useful texture features to classify images for the purpose of classifying different tumors.In this paper,Gabor wavelet transform is used to extract the texture features of MR images of pediatric posterior fossa tumors,and the statistical features are used to screen the extracted features.The useful feature vectors are selected and the support vector machine is used to train and classify the features to achieve the purpose of the classification of different tumors.The main work of this article is:In this paper,the method of image texture feature extraction and classification is studied.Based on the characteristics of common pediatric posterior fossa tumors,the Gabor wavelet transform and support vector machine are mainly studied.Gabor wavelet transform has the characteristics of multi-scale and multi-angle,and can obtain more abundant texture information than other texture analysis methods.Too many scales and angles will result in redundant information,increase the burden of follow-up training classification,and even interfere with the results.But too few scales and angles are selected to display sufficient texture information,showing no advantage of multiple scales and multiple angles.The appropriate Gabor parameters were selected through comparison of different scales and angles.The extracted eigenvectors are mean,con,ent,and asm in the angular direction.The four extracted feature vectors were analyzed using SPSS,and statistically different feature vectors were selected.And they are input into the support vector machine for training classification.In this paper,Gabor wavelet transform was performed on 22 ependymomas,23 medulloblastomas and astrocytomas from the First Affiliated Hospital of Zhengzhou University.According to the above method,four sets of feature vectors are extracted and texture analysis is performed.After the Gabor wavelet transform,the image has an edge effect.It will cause errors in the classification results.In this paper,based on the use of Gabor wavelet transform,the mathematical morphology method is used to perform a series of morphological operations on the image,eliminating the edge effect.The final experimental results show that the texture feature analysis based on Gabor filter can effectively achieve the classification of ependymomas and medulloblastomas and astrocytomas in children with posterior fossa tumors.It Can be used as a method of clinical diagnosis.