1.陆军军医大学(第三军医大学)第一附属医院放射科,重庆
2.推想医疗科技股份有限公司,北京
3.陆军军医大学大坪医院放射科,重庆
4.东部战区总医院放射诊断科,江苏南京
WANG Xianqi,E-mail:wxq9401@tmmu.edu.cn
Received:11 September 2025,
Revised:2025-11-21,
Accepted:25 November 2025,
Published:15 December 2025
移动端阅览
杨涛涛,陈伟,陈灿灿,等.融合深度学习、临床与影像组学特征的综合模型提升非小细胞肺癌EGFR突变预测效能[J].陆军军医大学学报,2025,47(23):2991-3001.DOI:10.16016/j.2097-0927.202509040.
YANG T T,CHEN W,CHEN C C,et al.Integrative model combining deep learning, clinical and radiomic features enhances EGFR mutation prediction in non-small cell lung cancer[J].J Army Med Univ,2025,47(23):2991-3001.DOI:10.16016/j.2097-0927.202509040.
杨涛涛,陈伟,陈灿灿,等.融合深度学习、临床与影像组学特征的综合模型提升非小细胞肺癌EGFR突变预测效能[J].陆军军医大学学报,2025,47(23):2991-3001.DOI:10.16016/j.2097-0927.202509040. DOI:
YANG T T,CHEN W,CHEN C C,et al.Integrative model combining deep learning, clinical and radiomic features enhances EGFR mutation prediction in non-small cell lung cancer[J].J Army Med Univ,2025,47(23):2991-3001.DOI:10.16016/j.2097-0927.202509040. DOI:
目的
2
探究胸部CT图像深度学习特征及联合临床特征、影像组学特征预测非小细胞肺癌(non-small cell lung cancer, NSCLC)表皮生长因子受体(epidermal growth factor receptor, EGFR)基因突变的应用价值。
方法
2
采用病例对照研究的方法,回顾性分析2013年1月至2023年10月3家医院放射科1 070例NSCLC患者的临床及影像资料。陆军军医大学第一附属医院719例NSCLC患者按照7∶3的比例随机分成训练集(
n
=502)、内部验证集(
n
=217);东部战区总医院173例患者作为外部验证集1;陆军军医大学大坪医院178例患者作为外部验证集2。深度学习特征通过使用主干为ResNet101的2.5D卷积神经网络(convolutional neural network, CNN)结构所获得,从基于CT的图像中提取影像组学特征,确定临床风险因素以构建临床预测模型,最后将深度学习特征、临床特征、影像组学特征整合构建综合模型。四种分类模型均采用随机森林(random forest, RF)的机器学习方法进行建模。绘制校正曲线以评估综合模型的拟合优度,绘制决策曲线评估模型的临床应用价值。
结果
2
深度学习模型在内部验证集、外部验证集1、外部验证集2中的曲线下面积(area under the curve, AUC)值分别为0.833 7(0.770 6~0.884 7)、0.815 1(0.741 6~0.882 8)和0.810 1(0.745 2~0.873 6),临床模型对应的AUC值分别为0.731 0(0.660 2~0.802 1)、0.746 0(0.666 4~0.824 9)和0.813 4(0.743 1~0.883 6),影像组学模型对应AUC值分别为0.762 4(0.692 4~0.825 1)、0.745 4(0.671 1~0.814 3)和0.724 7(0.639 7~0.801 6),整合后的综合模型预测效能值最优,对应AUC值分别为0.905 5(0.857 0~0.945 4)、0.832 7(0.763 3~0.896 4)和0.889 0(0.834 4~0.934 3)。决策曲线分析(decision curve analysis, DCA)表明,当阈值概率为0.15~0.85时,综合模型对EGFR的预测具有较高的净获益。
结论
2
基于CT图像的深度学习特征能较好地预测NSCLC患者的EGFR突变状态,当其联合临床特征、影像组学特征构建的综合模型可适当提高其预测效能。
Objective
2
To evaluate the predictive value of deep learning features from chest CT images combined with clinical and radiomics features for epidermal growth factor receptor (EGFR) mutations in non-small cell lung cancer (NSCLC).
Methods
2
This case-control study retrospectively analyzed clinical and imaging data of 1 070 NSCLC patients from radiology departments at three hospitals (January 2013 to October 2023). Patients were divided into: a training set (
n
=502) and internal validation set (
n
=217)
via
7∶3 randomization of 719 cases from the First Affiliated Hospital of Army Medical University; external validation set 1 (
n
=173) from General Hospital of Eastern Theater Command; external validation set 2 (
n
=178) from Daping Hospital of Army Medical University. Deep learning features were extracted using a 2.5D convolutional neural network (CNN) with ResNet101 backbone
radiomics features were derived from CT images
and clinical risk factors were identified to construct models. An integrated model combined deep learning
clinical
and radiomics features. All four models were developed using random forest (RF) classifiers. Calibration curves assessed goodness-of-fit
and decision curve analysis (DCA) evaluated clinical utility.
Results
2
The deep learning model achieved AUCs of 0.833 7 (95%
CI
: 0.770 6~0.884 7)
0.815 1 (0.741 6~0.882 8)
and 0.810 1 (0.745 2~0.873 6) in the internal and two external validation sets
respectively. Clinical models yielded AUCs of 0.731 0 (0.660 2~0.802 1)
0.746 0 (0.666 4~0.824 9)
and 0.813 4 (0.743 1~0.883 6); radiomics models showed AUCs of 0.762 4 (0.692 4~0.825 1)
0.745 4 (0.671 1~0.814 3)
and 0.724 7 (0.639 7~0.801 6). The integrated model demonstrated optimal performance with AUCs of 0.905 5 (0.857 0~0.945 4)
0.832 7 (0.763 3~0.896 4)
and 0.889 0 (0.834 4~0.934 3). DCA indicated significant net benefit for EGFR prediction at threshold probabilities of 0.15~0.85 using the integrated model.
Conclusion
2
Deep learning features from CT images effectively predict EGFR mutation status in NSCLC. The integrated model combining deep learning
clinical
and radiomics features further enhances predictive performance.
BRAY F , LAVERSANNE M , SUNG H , et al . Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries [J]. CA Cancer J Clin , 2024 , 74 ( 3 ): 229 - 263 . DOI: 10.3322/caac.21834 http://dx.doi.org/10.3322/caac.21834 .
URAS I Z , MOLL H P , CASANOVA E . Targeting KRAS mutant non-small-cell lung cancer: past, present and future [J]. Int J Mol Sci , 2020 , 21 ( 12 ): 4325 . DOI: 10.3390/ijms21124325 http://dx.doi.org/10.3390/ijms21124325 .
PASSARO A , MOK T , PETERS S , et al . Recent advances on the role of EGFR tyrosine kinase inhibitors in the management of NSCLC with uncommon, non exon 20 insertions, EGFR mutations [J]. J Thorac Oncol , 2021 , 16 ( 5 ): 764 - 773 . DOI: 10.1016/j.jtho.2020.12.002 http://dx.doi.org/10.1016/j.jtho.2020.12.002 .
PASCHE B , GRANT S C . Non-small cell lung cancer and precision medicine: a model for the incorporation of genomic features into clinical trial design [J]. JAMA , 2014 , 311 ( 19 ): 1975 - 1976 . DOI: 10.1001/jama.2014.3742 http://dx.doi.org/10.1001/jama.2014.3742 .
LEDUC C , MERLIO J P , BESSE B , et al . Clinical and molecular characteristics of non-small-cell lung cancer (NSCLC) harboring EGFR mutation: results of the nationwide French Cooperative Thoracic Intergroup (IFCT) program [J]. Ann Oncol , 2017 , 28 ( 11 ): 2715 - 2724 . DOI: 10.1093/annonc/mdx404 http://dx.doi.org/10.1093/annonc/mdx404 .
ZHOU Q , ZHANG X C , CHEN Z H , et al . Relative abundance of EGFR mutations predicts benefit from gefitinib treatment for advanced non-small-cell lung cancer [J]. J Clin Oncol , 2011 , 29 ( 24 ): 3316 - 3321 . DOI: 10.1200/JCO.2010.33.3757 http://dx.doi.org/10.1200/JCO.2010.33.3757 .
LINDEMAN N I , CAGLE P T , BEASLEY M B , et al . Molecular testing guideline for selection of lung cancer patients for EGFR and ALK tyrosine kinase inhibitors: guideline from the College of American Pathologists, International Association for the Study of Lung Cancer, and Association for Molecular Pathology [J]. J Thorac Oncol , 2013 , 8 ( 7 ): 823 - 859 . DOI: 10.1097/JTO.0b013e318290868f http://dx.doi.org/10.1097/JTO.0b013e318290868f .
AGGARWAL C , ROLFO C D , OXNARD G R , et al . Strategies for the successful implementation of plasma-based NSCLC genotyping in clinical practice [J]. Nat Rev Clin Oncol , 2021 , 18 ( 1 ): 56 - 62 . DOI: 10.1038/s41571-020-0423-x http://dx.doi.org/10.1038/s41571-020-0423-x .
ZITO MARINO F , BIANCO R , ACCARDO M , et al . Molecular heterogeneity in lung cancer: from mechanisms of origin to clinical implications [J]. Int J Med Sci , 2019 , 16 ( 7 ): 981 - 989 . DOI: 10.7150/ijms.34739 http://dx.doi.org/10.7150/ijms.34739 .
REUBEN A , ZHANG J X , CHIOU S H , et al . Comprehensive T cell repertoire characterization of non-small cell lung cancer [J]. Nat Commun , 2020 , 11 ( 1 ): 603 . DOI: 10.1038/s41467-019-14273-0 http://dx.doi.org/10.1038/s41467-019-14273-0 .
DETTERBECK F C , WOODARD G A , BADER A S , et al . The proposed ninth edition TNM classification of lung cancer [J]. Chest , 2024 , 166 ( 4 ): 882 - 895 . DOI: 10.1016/j.chest.2024.05.026 http://dx.doi.org/10.1016/j.chest.2024.05.026 .
SUCCONY L , RASSL D M , BARKER A P , et al . Adenocarcinoma spectrum lesions of the lung: Detection, pathology and treatment strategies [J]. Cancer Treat Rev , 2021 , 99 : 102237 . DOI: 10.1016/j.ctrv.2021.102237 http://dx.doi.org/10.1016/j.ctrv.2021.102237 .
LAMBIN P , RIOS-VELAZQUEZ E , LEIJENAAR R , et al . Radiomics: extracting more information from medical images using advanced feature analysis [J]. Eur J Cancer , 2012 , 48 ( 4 ): 441 - 446 . DOI: 10.1016/j.ejca.2011.11.036 http://dx.doi.org/10.1016/j.ejca.2011.11.036 .
HEILBRONER S P , MIOTTO R . Deep learning in medicine [J]. Clin J Am Soc Nephrol , 2023 , 18 ( 3 ): 397 - 399 . DOI: 10.2215/cjn.0000000000000080 http://dx.doi.org/10.2215/cjn.0000000000000080 .
CAI Z P , XU Z Y , CHEN Y F , et al . Multiparametric MRI subregion radiomics for preoperative assessment of high-risk subregions in microsatellite instability of rectal cancer patients: a multicenter study [J]. Int J Surg , 2024 , 110 ( 7 ): 4310 - 4319 . DOI: 10.1097/JS9.0000000000001335 http://dx.doi.org/10.1097/JS9.0000000000001335 .
WARKENTIN M T , AL-SAWAIHEY H , LAM S , et al . Radiomics analysis to predict pulmonary nodule malignancy using machine learning approaches [J]. Thorax , 2024 , 79 ( 4 ): 307 - 315 . DOI: 10.1136/thorax-2023-220226 http://dx.doi.org/10.1136/thorax-2023-220226 .
SHE Y L , HE B X , WANG F , et al . Deep learning for predicting major pathological response to neoadjuvant chemoimmunotherapy in non-small cell lung cancer: a multicentre study [J]. EBioMedicine , 2022 , 86 : 104364 . DOI: 10.1016/j.ebiom.2022.104364 http://dx.doi.org/10.1016/j.ebiom.2022.104364 .
YANG T T , WANG X Q , JIN Y , et al . Deep learning radiopathomics predicts targeted therapy sensitivity in EGFR-mutant lung adenocarcinoma [J]. J Transl Med , 2025 , 23 ( 1 ): 482 . DOI: 10.1186/s12967-025-06480-9 http://dx.doi.org/10.1186/s12967-025-06480-9 .
ROSSI G , BARABINO E , FEDELI A , et al . Radiomic detection of EGFR mutations in NSCLC [J]. Cancer Res , 2021 , 81 ( 3 ): 724 - 731 . DOI: 10.1158/0008-5472.CAN-20-0999 http://dx.doi.org/10.1158/0008-5472.CAN-20-0999 .
WANG S , SHI J Y , YE Z X , et al . Predicting EGFR mutation status in lung adenocarcinoma on computed tomography image using deep learning [J]. Eur Respir J , 2019 , 53 ( 3 ): 1800986 . DOI: 10.1183/13993003.00986-2018 http://dx.doi.org/10.1183/13993003.00986-2018 .
LAN T J , KUANG S J , LIANG P S , et al . MRI-based deep learning and radiomics for prediction of occult cervical lymph node metastasis and prognosis in early-stage oral and oropharyngeal squamous cell carcinoma: a diagnostic study [J]. Int J Surg , 2024 , 110 ( 8 ): 4648 - 4659 . DOI: 10.1097/JS9.0000000000001578 http://dx.doi.org/10.1097/JS9.0000000000001578 .
SHI S Y , LIN C X , ZHOU J , et al . Development and validation of a deep learning radiomics model with clinical-radiological characteristics for the identification of occult peritoneal metastases in patients with pancreatic ductal adenocarcinoma [J]. Int J Surg , 2024 , 110 ( 5 ): 2669 - 2678 . DOI: 10.1097/JS9.0000000000001213 http://dx.doi.org/10.1097/JS9.0000000000001213 .
GIRARD N , SIMA C S , JACKMAN D M , et al . Nomogram to predict the presence of EGFR activating mutation in lung adenocarcinoma [J]. Eur Respir J , 2012 , 39 ( 2 ): 366 - 372 . DOI: 10.1183/09031936.00010111 http://dx.doi.org/10.1183/09031936.00010111 .
DU W X , QIU T , LIU H Q , et al . The predictive value of serum tumor markers for EGFR mutation in non-small cell lung cancer patients with non-stage IA [J]. Heliyon , 2024 , 10 ( 9 ): e29605 . DOI: 10.1016/j.heliyon.2024.e29605 http://dx.doi.org/10.1016/j.heliyon.2024.e29605 .
杨涛涛 , 王显棋 , 陈灿灿 , 等 . 基于CT影像组学特征联合临床特征预测非小细胞肺癌EGFR突变状态 [J]. 陆军军医大学学报 , 2025 , 47 ( 8 ): 847 - 857 . DOI: 10.16016/j.2097-0927.202411024 http://dx.doi.org/10.16016/j.2097-0927.202411024 .
YANG T T , WANG X Q , CHEN C C , et al . Prediction of EGFR mutation status in non-small cell lung cancer based on CT radiomic features combined with clinical characteristics [J]. J Army Med Univ , 2025 , 47 ( 8 ): 847 - 857 . DOI: 10.16016/j.2097-0927.202411024 http://dx.doi.org/10.16016/j.2097-0927.202411024 .
YATABE Y , KERR K M , UTOMO A , et al . EGFR mutation testing practices within the Asia Pacific region: results of a multicenter diagnostic survey [J]. J Thorac Oncol , 2015 , 10 ( 3 ): 438 - 445 . DOI: 10.1097/JTO. 0000000000000422 http://dx.doi.org/10.1097/JTO.0000000000000422 .
QI Y J , SU G H , YOU C , et al . Radiomics in breast cancer: Current advances and future directions [J]. Cell Rep Med , 2024 , 5 ( 9 ): 101719 . DOI: 10.1016/j.xcrm.2024. 101719 http://dx.doi.org/10.1016/j.xcrm.2024.101719 .
DU Y , CAI M J , ZHA H L , et al . Ultrasound radiomics-based nomogram to predict lymphovascular invasion in invasive breast cancer: a multicenter, retrospective study [J]. Eur Radiol , 2024 , 34 ( 1 ): 136 - 148 . DOI: 10.1007/s00330-023-09995-1 http://dx.doi.org/10.1007/s00330-023-09995-1 .
邓冰 , 黄文才 , 江远亮 . 人工智能和影像组学在肺癌精准诊疗中的应用进展 [J]. 重庆医学 , 2025 , 54 ( 1 ): 213 - 218 . DOI: 10.3969/j.issn.1671-8348.2025.01.039 http://dx.doi.org/10.3969/j.issn.1671-8348.2025.01.039 .
DENG B , HUANG W C , JIANG Y L . Progress in application of artificial intelligence and radiomics in precision diagnosis and treatment of lung cancer [J]. Chongqing Med , 2025 , 54 ( 1 ): 213 - 218 . DOI: 10.3969/j.issn.1671-8348.2025.01.039 http://dx.doi.org/10.3969/j.issn.1671-8348.2025.01.039 .
LIU Y , KIM J , BALAGURUNATHAN Y , et al . Radiomic features are associated with EGFR mutation status in lung adenocarcinomas [J]. Clin Lung Cancer , 2016 , 17 ( 5 ): 441 - 448.e6 . DOI: 10.1016/j.cllc.2016.02.001 http://dx.doi.org/10.1016/j.cllc.2016.02.001 .
WANG S , YU H , GAN Y C , et al . Mining whole-lung information by artificial intelligence for predicting EGFR genotype and targeted therapy response in lung cancer: a multicohort study [J]. Lancet Digit Health , 2022 , 4 ( 5 ): e309 - e319 . DOI: 10.1016/S2589-7500(22)00024-3 http://dx.doi.org/10.1016/S2589-7500(22)00024-3 .
WANG X Y , WU S H , REN J , et al . Predicting gene comutation of EGFR and TP53 by radiomics and deep learning in patients with lung adenocarcinomas [J]. J Thorac Imaging , 2025 , 40 ( 3 ): e0817 . DOI: 10.1097/RTI. 0000000000000817 http://dx.doi.org/10.1097/RTI.0000000000000817 .
MAHAJAN A , KANIA V , AGARWAL U , et al . Deep-learning-based predictive imaging biomarker model for EGFR mutation status in non-small cell lung cancer from CT imaging [J]. Cancers , 2024 , 16 ( 6 ): 1130 . DOI: 10.3390/cancers16061130 http://dx.doi.org/10.3390/cancers16061130 .
0
Views
0
下载量
0
CSCD
Publicity Resources
Related Articles
Related Author
Related Institution
京公网安备11010802024621
