A Preliminary Study On The Application Of Radiomics In The Target Delineation Of Nasopharyngeal Carcinoma

Radiomics is a new computer-aided image processing and diagnosis tool,which is mainly used in clinical prognosis prediction,tumor microenvironment research,combined with pathological examination and other purposes.As a result of the development of imaging examination and radiotherapy technology,individualized radiotherapy schemes have become possible,giving birth to a large number of advanced image processing and analysis techniques.Radiomics has been widely studied because of its ability in mine the data information contained in the image.This paper describes the application of radiomics in solving some existing problems in individualized radiotherapy for nasopharyngeal carcinoma(NPC).Mainly soluted or interpretated the problems of differentiation between tumor and adjacent tissue,segmentation of different heterogeneous regions such as hypoxia in tumor,identification of small size metastatic lymph nodes.It has certain reference value in perfecting individualized radiotherapy for NPC.This paper mainly introduces the research methods and technical characteristics of radiomics,and applies it in the correlation analysis of NPC.In this paper,a total of 108 patients with NPC were selected,and the data of CT,MRI,PET and pathological examination were collected.In previous studies,such as quantitative analysis of tumor and muscle tissue,distinction between hypoxic and non-hypoxic regions in tumor,statistical methods were used to obtain features of practical value.Machine learning method was used to build model and identify the data in the identification of small volume cervical metastatic lymph nodes.The main results were as follows:(1)In the study of quantitative analysis of tumor and muscle tissue,some radiomics features obtained in CT images changed with image enhancement states,which can reflect the dynamic changes of tumor heterogeneity to quantify the difference between tumor and normal tissue.(2)In the study of distinguishing hypoxic and non-hypoxic regions in tumors,there were 4 features changed more than 90% in the MRI T1,T1+ and T2 images,which could reflect the changes of the hypoxic region of the tumor target area in NPC.Quantifying and tracking the changes of these features may helpful to identify the hypoxic area of NPC.(3)In the study of differentiating cervical metastatic lymph nodes.The machine learning prediction model,which is composed of CT imaging macroscopic indexes and radiomics features,can accurately predict metastatic lymph nodes of NPC with accuracy of 91.7%,which provided a certain reference for clinical identification of metastatic lymph nodes at small volume.The results have certain reference value in tumor segmentation,hypoxic region identification and metastatic lymph node identification of NPC,may beneficial to implement individualized treatment schemes such as different prescription doses or simultaneous dose radiotherapy.