| Objective By prospectively selected 92 patients with lung adenocarcinoma and collected their clinical data and MRI morphological features,the clinical predictors and MRI morphological features related to the EGFR gene mutation status of lung adenocarcinoma were screened.To compare the value of radiomics model based on T2WI,DWI,ADC and nomogram of combined clinical factors in preoperative prediction of EGFR mutation status in lung adenocarcinoma.Methods A total of 92 patients with lung adenocarcinoma confirmed by pathology from November 2015 to September 2019 were prospectively selected,including 53 males and 39 females,aged from 27 to 79 years old,with an average age of 57.65±10.90 years old.There were 51 cases with EGFR mutation and 41 cases without mutation.Clinical stages:I stage 34 cases,II stage 14 cases,III stage 20 cases,IV stage 24 cases.All patients underwent chest MRI scan within one week before surgery,including T2WI sequence,EPI-DWI(b=0,800s/mm~2)sequence,and generated corresponding ADC images.The clinical data of 92 patients were age,sex,smoking history,clinical stage,CEA,CA125 and CA153 within 1 week before surgery.The MRI morphological features of 92 patients were:tumor location,maximum diameter of tumors,lobular sign,spicule sign,pleural indentation sign and ADC value of tumors.SPSS statistical software was used to analyze the clin-ical data and imaging signs of patients with EGFR mutation and EGFR nonmutation,and to screen the clinical factors and MRI imaging signs related to the mutation status of EGFR gene in lung adenocarcinoma.Constructed an radiomics models based on multi-modal MRI-based:Using software ITK-SNAP in T2WI,DWI and ADC images,manually delineate and segment all levels of all lesions,ROI is placed along the outline of the le-sions,and import the original images and ROI files of all patients into GE AK software for image preprocessing.Six categories of radiomics features,namely Histogram,Form Factor the Features,gray level co-occurrence matrix(GLCM),Haralick parameters,graylevel run-length matrix(GLRLM)and Gray Level Size Zone Matrix(GLSZM),were selected in AK software for feature extraction.A total of 396 radiomics features were extracted from each sequence of each patient.The data of 92 patients were randomly divided into the training group(65 cases)and the test group(27 cases)at a ratio of 7:3 by R software.The extracted 3D radiomics features were statistically analyzed and the corresponding statistical test was completed.The least absolute shrinkage and selection operator(LASSO)regression was used to reduce the dimension of radiomics features,and the radiomics signatures of ADC,DWI,T2WI and merged sequences were constructed respectively.The combined prediction model of multi-factor logistic regression was established by radiomics signatures of merged sequences and clinical predictors,and the nomogram was made.The area under the curve(AUC)was used to evaluate the predictive efficacy of five radiomics models on EGFR mutation in lung adenocarcinoma.DeLong test was used to compare whether the predictive efficacy of the five models was statistically different(P<0.05 was considered statistically significant).The decision curve analysis was used to evaluate the benefits of different threshold probabilities of each model.Calibration curve is used to evaluate the predictive efficiency of nomogram.Results Among the 92 patients with lung adenocarcinoma,53 were males and 39were females,aged from 27 to 79 years old,with an average age of 57.65±10.90 years old.There were 51 cases of EGFR mutation and 41 cases of EGFR nonmutation.Clinical staging:34 cases of stage I,14 cases of stage II,20 cases of stage III and 24 cases of stage IV.There were 28 cases with smoking history and 64 cases without smoking history.The CEA:0.27?3101 ng/ml,with an average value 106.06±412.86 ng/ml;CA125:2.57?2014U/ml,with an average value 86.57±248.13U/ml;CA153:4.57?181.60U/ml,with an average value of 27.32±30.80U/ml.MRI morphological features of92 patients,27 cases in the right upper lung,10 cases in the right middle lung,18 cases in the right lower lung,24 in the left upper lung,and 13 in the left lower lung,The maximum tumor diameter range was 1.02?8.93cm,with an average of 3.90±1.99cm.There were 53cases without lobular sign,39 cases with lobular sign,67 cases without spicule sign,25cases with spicule sign,68 cases without pleural indentation sign,and 24 cases with pleu-ral indentation sign.ADC values ranged from 603 to 2335mm~2/s,with an average of1171.59±286.67mm~2/s.There were statistically significant differences in gender,smoking status and ADC values between the EGFR mutant and nonmutant groups(P<0.05).There were 65 cases in the training group,including 39 cases of EGFR mutation and 26 cases of EGFR nonmutation.There were statistical differences in gender and smoking status in the training group(P<0.05).There were 27 patients in the test group,including 12 patients with EGFR mutation and 15 patients without EGFR nonmutation.There were statistical differences in gender and smoking status of the patients in the test group(P<0.05).Gender and smoking status were screened as clinical factors for predicting EGFR mutation status in lung adenocarcinoma.After screening the features extracted from ADC,DWI,T2WI and combined sequences,the remaining 3,6,4 and 6 radiomics features were left,respectively.The AUC of the training group and the test group in the ADC model,DWI model,T2WI model and combined model were 0.874(95%CI:0.786 to 0.962)and 0.806(95%CI:0.611 to 1.000),0.784(95%CI:0.675 to 0.893)and 0.722(95%CI:0.520 to 0.925),0.692(95%CI:0.563 to 0.821)and 0.656(95%CI:0.439 to 0.873),0.889(95%CI:0.809 to0.968)and 0.839(95%CI:0.685 to 0.993).The AUC of the combined model with gender,smoking status and radiomics signatures in the training group was 0.925(95%CI:0.856 to0.994),with sensitivity 84.6%,specificity 96.2%and accuracy 89.2%.The AUC of the test group was 0.728(95%CI:0.531 to 0.925),the sensitivity was 66.7%,the specificity was 73.3%,and the accuracy was 70.4%.Conclusion Gender and smoking status can be used as clinical factors to predict EGFR mutation status in lung adenocarcinoma.ADC and DWI model were slightly better than T2WI model,and the prediction performance of multisequence radiomics model was further improved than that of single-sequence model.The nomogram based on MRI radiomics signatures and clinical factors can be used as a quantitative tool to predict the EGFR mutation status in preoperative lung adenocarcinoma. |
PDF Full Text Request