A Study Of Serum Marker Combined With Texture Analysis Based On CT Images Predict The Radio-Chemotherapy Sensitivity Of Local Advanced Non-Small-Cell Lung Cancer

Purpose: To investigate the feasibility of predicting the treatment response of local advanced non-small cell lung cancer(NSCLC)patients to radio-chemotherapy by serum markers combined with the texture analysis based on computed tomography(CT)images.Methods: This study retrospectively enrolled 106 patients with pathologically confirmed stage III NSCLC who underwent concurrent chemoradiotherapy in our hospital from June2014 to June 2017.Collecting pre-treatment serum markers(including the count of red blood cell,hemoglobin,platelet and albumin,the tumor marker of NSE,Cyfra21-1 and CEA,the count of lymphocytes,neutrophils and monocytes,percentage of lymphocytes,neutrophils and monocytes,and NLR,PLR,LMR,NMR,PMR)and clinical datum(including age,gender,smoking history,KPS score,T stage,N stage,clinical stage,pathological type,lesions),and enhanced CT images in a week before treatment.The efficacy was evaluated according to the RECIST 1.1 criteria.The PR and CR patients were divided into the radio-chemotherapy-sensitive group,while the SD and PD patients were divided into the radio-chemotherapy-resistant group.The region of interest(ROI)was delineated at each target level of the tumor,and 3D image reconstruction was performed.The 3D texture analysis was performed on the image by MATLAB software to extract 3D texture parameters.Filtering out statistically significant texture parameters by single factor and multi-factor analysis,to establish a serum marker combined with texture analysis prediction model.Results: Based on the RECIST 1.1 criteria,64(PR+CR)patients were classified as responders,while 14(SD+PD)were classified as nonresponders.The chi-square test showed that the clinical factors N stage(P=0.025)and the serum marker Cyfra21-1(P=0.009)were correlated with the sensitivity of radio-chemotherapy.The difference between the two groups was statistically significant.According to the results of the independent-sample T test,2texture parameters extracted from 3D images had significant difference between the twogroups.Internal validation was performed using stratified 5-fold cross-validation,suggesting that the parameters are reliable.Using logistic regression multivariate analysis,Cyfra21-1(P=0.037)combined with Cluster Tendency Range(P= 0.047)was statistically significant.Conclusions: Both of the serum markers and the texture parameters extracted from CT images could act as a predictive model of tumor response to radio-chemotherapy in local advanced NSCLC patients.