Peritumoral Expansion-Based CT Radiomics for Predicting EGFR Mutation Status in Non-Small Cell Lung Cancer
10.3969/j.issn.1005-5185.2025.11.005
- VernacularTitle:基于瘤周扩展CT影像组学模型预测非小细胞肺癌EGFR突变状态
- Author:
Xiaoyan WANG
1
;
Zhicheng ZHANG
;
Yan ZENG
;
Lili GUO
Author Information
1. 南京医科大学附属淮安第一医院影像科,江苏 淮安 223300
- Publication Type:Journal Article
- Keywords:
Carcinoma,non-small-cell lung;
Tomography,X-ray computed;
Radiomics;
Epidermal growth factor receptor;
Mutation;
Forecasting;
Peritumoral
- From:
Chinese Journal of Medical Imaging
2025;33(11):1164-1172
- CountryChina
- Language:Chinese
-
Abstract:
Purpose To investigate the value of peritumoral expansion-based radiomics from tumor regions of interest(ROIs)for predicting epidermal growth factor receptor(EGFR)mutation status and to identify the optimal peritumoral expansion margin.Materials and Methods This retrospective study included 390 patients with pathologically confirmed non-small cell lung cancer(NSCLC)and definitive EGFR genotyping results from the Affiliated Huai'an No.1 People's Hospital of Nanjing Medical University(December 2021 to September 2023).Patients were randomly divided into training and validation sets(8∶2 ratio,310 vs.80 patients;CT slice thickness:5-7 mm).An additional independent test set of 100 patients undergoing thin-section CT(1-2 mm)was included to assess model generalizability across slice thicknesses.Clinical characteristics and CT semantic features were analyzed.After automated tumor segmentation with manual refinement,ROIs were sequentially expanded outward by 1,2,3,4 and 5 mm.A total of 2 264 radiomic features were extracted from each ROI.Feature selection using minimum redundancy maximum relevance and least absolute shrinkage and selection operator algorithms was performed in the training set to calculate radiomics scores.Logistic regression was used to develop prediction models.A combined model integrating optimal peritumoral radiomics score with clinical and CT features was established,with performance compared across models.Results Gender(χ2=24.922,P<0.001),smoking history(χ2=11.199,P=0.001),emphysema(χ2=40.802,P<0.001),and lymph node metastasis(χ2=5.674,P=0.017)were associated with EGFR mutation status.In the validation set,the area under the curve for the tumor ROI-based radiomics model was 0.676,while the five expansion models achieved area under the curve of 0.723,0.720,0.734,0.681,and 0.598,respectively.The combined model based on 3 mm expansion demonstrated superior performance to individual radiomics and clinical models,with area under the curve of 0.826,0.734,0.711 in the validation set,and 0.809,0.760,0.702 in the independent test set.Conclusion Peritumoral expansion-based CT radiomics demonstrates value for predicting EGFR mutations in NSCLC,with 3 mm identified as the optimal expansion margin.Integration with clinical information further enhances predictive accuracy.
