Predicting Mutations Of The Epidermal Growth Factor Receptor Gene In Non-small Cell Lung Cancer By Quantitative Radiomic Features

At present,finding a method to cure lung cancer is still a difficult part of global medical technology,and how to decrease the morbidity and mortality also is a thorny problem.In consideration of non-small cell lung cancer(NSCLC)account for 85% in lung cancer,NSCLC is the focus and interest of current research.With the development of medical level,many studies have focused on genes in NSCLC,in which the detection of epidermal growth factor receptor(EGFR)gene mutation has important clinical significance for the treatment of patients and the improvement of prognosis.Targeted therapy of tyrosine kinase inhibitors(TKI)for treatment has become a widely accepted method when patients with advanced NSCLC are specifically diagnosed as EGFR mutations.Amplifcation Refractory Mutation System(ARMS)and direct sequencing are the current reliable method to exam the mutation of exon for the EGFR.However,these pervasive approaches are available on the condition of surgery and latest technology,which need spend much money and wait for long time.Furthermore,the patients with EGFR often belong to locally advanced or late diagnosis,if you do not take the treatment,the survival rate less than 10%.According to the clinical problem,we presented radiomic method,an emerging field potentially to link the medical images with the genomic characteristics of human tumors by extracting many features from images,provides chances for non-invasive diagnostics and prognostics rather than the surgery.Therefore,we want to employ radiomics to discover the relationship between EGFR mutation status and the radiomic features.In order to accomplish this,Here we introduce a 495 raidomic features extracted form 180 cases of NSCLC with CT scans to analyze the heterogeneity and phenotype of region of interest(ROI)of tumor segmented by the experienced radiologist.We utlilize least absolute shrinkage and selection operator(Lasso)method to selected the representative features to build the logistic regression model and analysis the performance by the receiver operating characteristic(ROC)curve and found that radiomics features play dominating role to predict the mutation of EGFR combined with clinical variables.Moreover,we depicted the nomogram to visually present the relationship between radiomic features and clinical features individually.This method may help radiologists better rely on the discriminative mineable findings in molecular phenotypic and to translate image information into clinical practice for disease diagnosis,which provides a new way for patients with nonsmall cell lung cancer to detect EGFR mutation.