Detection of epithelial growth factor receptor (EGFR) mutations on CT images of patients with lung adenocarcinoma using radiomics and/or multi-level residual convolutionary neural networks

Li, Xiaoyang; Xiong, Junfeng; Jia, Tian-Ying; Shen, Tongyi; Hou, Ronghui; Zhao, Jun; Fu, Xiaolong

doi:10.21037/jtd.2018.11.03

Cited by 31 publications

(39 citation statements)

References 36 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Previous articles on pulmonary tumor radiomics were generally based on non-contrast CT images (20)(21)(22). Some studies have used contrast images alone (23,24), and some have used both, but no comparisons or descriptions regarding which type of image is better for further analysis have been reported (25).…”

Section: Discussionmentioning

confidence: 99%

Radiomics Signature as a Predictive Factor for EGFR Mutations in Advanced Lung Adenocarcinoma

Hong

Zhang

et al. 2020

Front. Oncol.

View full text Add to dashboard Cite

Purpose: To develop and validate a radiomic signature to identify EGFR mutations in patients with advanced lung adenocarcinoma.Methods: This study involved 201 patients with advanced lung adenocarcinoma (140 in the training cohort and 61 in the validation cohort). A total of 396 features were extracted from manual segmentation based on enhanced and non-enhance CT imaging after image preprocessing. The Lasso algorithm was used for feature selection, 6 machine learning methods were used to construct radiomics models. Receiver operating characteristic (ROC) curve analysis was applied to evaluate the performance of the radiomic signature between different data and methods. A nomogram was developed using clinical factors and the radiomics signature, then it was analyzed based on its discriminatory ability and calibration. Decision curve analysis (DCA) was implemented to evaluate the clinical utility.Results: Ten features for contrast data and eleven features for non-contrast data were selected through LASSO algorithm. The performance of the radiomics signature for contrast images was better than that for non-contrast images in all of the 6 different machine learning methods. Finally, the best radiomics signature was built with logistic regression method based on enhanced CT imaging with an area under the curve (AUC) of 0.851 (95% CI, 0.750 to 0.951) in the validation cohort. A nomogram was developed using the radiomics signature and sex with a C-index of 0.908 (95% CI, 0.862 to 0.954) in the training cohort and 0.835 (95% CI, 0.825 to 0.845) in the validation cohort. It showed good discrimination and calibration (Hosmer-Lemeshow test, P = 0.621 for the training cohort and P = 0.605 for the validation cohort).Conclusion: Radiomics signature can help to distinguish between EGFR positive and wild type advanced lung adenocarcinomas.

show abstract

Section: Discussionmentioning

confidence: 99%

Radiomics Signature as a Predictive Factor for EGFR Mutations in Advanced Lung Adenocarcinoma

Hong

Zhang

et al. 2020

Front. Oncol.

View full text Add to dashboard Cite

show abstract

“…Although Liu et al [27] and Rizzo et al [29] achieved relatively higher AUC scores of 0.709 and 0.82, respectively, their studies did not include independent testing groups which are essential to establishing the robustness of predictive models. Xiong et al [23] and Li et al [19] showed impressive results by using 3D deep learning model trained from scratch and achieved AUC scores of 0.776 and 0.809, respectively. Our model was trained and validated using the same patient cohort as Li's work [19] and, in an independent testing group, achieved a higher AUC score than any previously published study in this area.…”

Section: Discussionmentioning

confidence: 99%

“…Xiong et al [23] and Li et al [19] showed impressive results by using 3D deep learning model trained from scratch and achieved AUC scores of 0.776 and 0.809, respectively. Our model was trained and validated using the same patient cohort as Li's work [19] and, in an independent testing group, achieved a higher AUC score than any previously published study in this area. Notably, the lowest-performing model in our study (a fine-tuned 2D CNN using the large input size and transverse slicing method) achieved a lower AUC (0.642) than any prior studies, highlighting the importance of optimally selecting the methods used to construct CNNs for medical applications.…”

Section: Discussionmentioning

confidence: 99%

“…As a result, there has been no 3D pre-trained CNN architecture from which a new 3D CNN model could be fine-tuned for a specific clinical application. One approach to construct a 3D clinical model is to crop volumes of interest (VOI) from full-size 3D image series [5], [18], [19] and then use the VOIs to train the 3D model from scratch. Such models may not be robust due to overfitting from the limited training sample data.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Implementation Strategy of a CNN Model Affects the Performance of CT Assessment of EGFR Mutation Status in Lung Cancer Patients

Xiong

Lin

et al. 2019

IEEE Access

Self Cite

View full text Add to dashboard Cite

To compare CNN models implemented using different strategies in the CT assessment of EGFR mutation status in patients with lung adenocarcinoma. 1,010 consecutive lung adenocarcinoma patients with known EGFR mutation status were randomly divided into a training set (n = 810) and a testing set (n = 200). The CNN models were constructed based on ResNet-101 architecture but implemented using different strategies: dimension filters (2D/3D), input sizes (small/middle/large and their fusion), slicing methods (transverse plane only and arbitrary multi-view planes), and training approaches (from scratch and finetuning a pre-trained CNN). The performance of the CNN models was compared using AUC. The fusion approach yielded consistently better performance than other input sizes, although the effect often did not reach statistical significance. Multi-view slicing was significantly superior to the transverse method when fine-tuning a pre-trained 2D CNN but not a CNN trained from scratch. The 3D CNN was significantly better than the 2D transverse plane method but only marginally better than the multi-view slicing method when trained from scratch. The highest performance (AUC = 0.838) was achieved for the fine-tuned 2D CNN model when built using the fusion input size and multi-view slicing method. The assessment of EGFR mutation status in patients is more accurate when CNN models use more spatial information and are finetuned by transfer learning. Our finding of the implementation strategy of a CNN model could be a guide to other medical 3D images applications. Compared with other published studies which used medical images to identify EGFR mutation status, our CNN model achieved the best performance in the biggest patient cohort.INDEX TERMS CNN, EGFR, implementation strategy.

show abstract

“…This approach is a non-invasive and easy-toimplement deep learning method. Other studies (27)(28)(29) also show that deep learning models can identify gene mutations in lung cancer.…”

Section: Original Articlementioning

confidence: 95%

Multi-channel multi-task deep learning for predicting EGFR and KRAS mutations of non-small cell lung cancer on CT images

Dong

Hou

Yang

et al. 2021

Quant Imaging Med Surg

View full text Add to dashboard Cite

Background: Predicting the mutation statuses of 2 essential pathogenic genes [epidermal growth factor receptor (EGFR) and Kirsten rat sarcoma (KRAS)] in non-small cell lung cancer (NSCLC) based on CT is valuable for targeted therapy because it is a non-invasive and less costly method. Although deep learning technology has realized substantial computer vision achievements, CT imaging being used to predict gene mutations remains challenging due to small dataset limitations. Methods:We propose a multi-channel and multi-task deep learning (MMDL) model for the simultaneous prediction of EGFR and KRAS mutation statuses based on CT images. First, we decomposed each 3D lung nodule into 9 views. Then, we used the pre-trained inception-attention-resnet model for each view to learn the features of the nodules. By combining 9 inception-attention-resnet models to predict the types of gene mutations in lung nodules, the models were adaptively weighted, and the proposed MMDL model could be trained end-to-end. The MMDL model utilized multiple channels to characterize the nodule more comprehensively and integrate patient personal information into our learning process. Results:We trained the proposed MMDL model using a dataset of 363 patients collected by our partner hospital and conducted a multi-center validation on 162 patients in The Cancer Imaging Archive (TCIA) public dataset. The accuracies for the prediction of EGFR and KRAS mutations were, respectively, 79.43% and 72.25% in the training dataset and 75.06% and 69.64% in the validation dataset. Conclusions:The experimental results demonstrated that the proposed MMDL model outperformed the latest methods in predicting EGFR and KRAS mutations in NSCLC.

show abstract

Detection of epithelial growth factor receptor (EGFR) mutations on CT images of patients with lung adenocarcinoma using radiomics and/or multi-level residual convolutionary neural networks

Cited by 31 publications

References 36 publications

Radiomics Signature as a Predictive Factor for EGFR Mutations in Advanced Lung Adenocarcinoma

Radiomics Signature as a Predictive Factor for EGFR Mutations in Advanced Lung Adenocarcinoma

Implementation Strategy of a CNN Model Affects the Performance of CT Assessment of EGFR Mutation Status in Lung Cancer Patients

Multi-channel multi-task deep learning for predicting EGFR and KRAS mutations of non-small cell lung cancer on CT images

Contact Info

Product

Resources

About