Habitat model based on lung CT for predicting brain metastasis of lung adenocarcinoma with epidermal growth factor receptor mutation
10.13929/j.issn.1672-8475.2024.07.006
- VernacularTitle:基于肺部CT生境模型预测表皮生长因子受体突变型肺腺癌脑转移
- Author:
Lijuan LIN
1
;
Ying LIN
;
Yanqing WU
;
Xiang LIN
;
Wei GUO
;
Yang SONG
;
Dehua CHEN
Author Information
1. 福建医科大学附属第一医院医学影像科,福建 福州 350005;福建医科大学附属第一医院滨海院区国家区域医疗中心医学影像科,福建 福州 350212
- Keywords:
lung neoplasms;
brain neoplasms;
adenocarcinoma;
receptor,epidermal growth factor;
tomography,X-ray computed;
radiomics
- From:
Chinese Journal of Interventional Imaging and Therapy
2024;21(7):408-413
- CountryChina
- Language:Chinese
-
Abstract:
Objective To observe the value of habitat model based on lung CT for predicting brain metastasis(BM)of lung adenocarcinoma with epidermal growth factor receptor(EGFR)mutation.Methods Data of plain lung CT of 198 lung adenocarcinoma patients with EGFR-mutant were retrospectively analyzed.The patients were divided into training set(n=138)and test set(n=60)at the ratio of 7∶3,and further divided into BM subgroup and non-BM subgroup in each set.Then a logistic regression(LR)clinical model was constructed using variables being statistically different between subgroups in training set.For features extracted from tumor and subregion of tumor,radiomics models and habitat models were constructed based on random forest,Gaussian process(GP)and support vector machine(SVM)algorithms,and the best radiomics and habitat models with generalization ability were screened.LR combined model was constructed based on the predicted values of the best radiomics and habitat models with generalization ability,as well as the clinical model.Then receiver operating characteristic curves were drawn,and the area under the curves(AUC)were calculated to evaluate the efficacy of each model for predicting BM of lung adenocarcinoma with EGFR-mutant.Spearman correlation analysis was performed to observe the correlations between Ki-67 and habitat features of lung adenocarcinoma with EGFR-mutant.Results AUC of LR clinical model,GP radiomics model,SVM habitat model and LR combined model for predicting BM of lung adenocarcinoma with EGFR-mutant was 0.700,0.726,0.801 and 0.834 in training set,0.754,0.600,0.715 and 0.848 in test set,respectively.AUC of LR combined model was higher than that of LR clinical model in training set(P＜0.001),also higher than that of GP radiomics model in test set(P=0.010).Compared with GP radiomics model and SVM habitat model,the performance of LR combined model was significantly and positively improved in training set(integrated discrimination improvement index[IDI]=8.60%,8.55%,both P＜0.001).Ki-67 level of EGFR-mutant lung adenocarcinoma was lowly and positively correlated with habitatmap_original_glszm_lalgle extracted from habitat map(│rs│=0.201,P=0.004).Conclusion The habitat model based on lung CT could be used to predict BM of lung adenocarcinoma with EGFR-mutant effectively.
