MR Imaging Radiomics Signatures For Predicting The Risk Of Distant Metastasis Of Nasopharyngeal Carcinoma

Nasopharyngeal carcinoma(NPC)is the most common histologic malignancy of nasopharynx.Because most of the patients are concentrated in Guangdong Province,nasopharyngeal carcinoma is also known as " Canton tumor ".Distant metastasis is one of the main reasons to lead to the death of the NPC patients.Therefore,predicting the risk of distant metastasis of nasopharyngeal carcinoma is an important key to improve the treatment outcomes and the survival rate of NPC patients.Medical imaging is an important field in medical research and has been widely applied to clinical treatment.By using medical imaging,we can observe the strong phenotypic differences of human cancers noninvasively.In practical clinical applications,there are many mature imaging technologies.In the field of treatment of nasopharyngeal carcinoma,magnetic resonance imaging(MRI)has great advantages.In the last three decades,medical imaging makes progress in imaging technology.In recent years,radiomics has become a new development direction in medical imaging.Radiomics refers to the comprehensive quantification of tumor phenotypes by applying a large number of quantitative image features.The research of radiomics is developing rapidly and has been applied to the diagnosis,treatment and prediction of some tumors.But the research and application of radiomics in nasopharyngeal carcinoma is almost blank.In this study,we analyzed the risk of distant metastasis of nasopharyngeal carcinoma based on magnetic resonance images.To achieve this goal,we presented a radiomics analysis of 431 features quantifying tumor image intensity,shape and texture,which are extracted from computed tomography data of 700 patients with nasopharyngeal carcinoma.Then,principal component analysis and unsupervised clustering were applied to analyze the correlation between radiomics data and clinical data.Finally,a risk prediction model was constructed by regression analysis.Based on the experimental results,we find that a large number of radiomics features are strongly associated with clinical data and have prognostic power in independent data sets of NPC patients,many of which were not identified as significant before.The results may have a clinical impact as imaging is routinely used in clinical practice,providing an unprecedented opportunity to improve decision-support in nasopharyngeal carcinoma treatment at low cost.