Application of MRI texture analysis in the study of the posterior fossa tumors growing trend in children.

Abstract

In order to analyze the growing trend of the posterior fossa tumor in children and provide assistant basis for the treatment or surgery of tumors, a variety of texture analysis methods were comprehensive used to analyze and identify three kinds of brain tissues, tumor region, tumor diffusion region and normal brain tissue region. The MRIs of tumor patients were collected to extract texture features. Then feature selection method CFS and feature compression method partial least squares regression (PLSR) were used to process these feature space. Finally, different classification methods were used to identify three classes samples expressed in different forms. The classification results of all features show that texture analysis can be used to analyze the growing trend of the tumor and provide sufficient support for the prediction of it. The CFS subsets results show that the specific texture features have important value for qualitative analysis and discrimination of three kinds of tissues. PLSR compressed sets results confirm the above results and provide intuitive display of compressed sample space distribution.