Three-Dimensional Convolutional Neural Network-Based Prediction of Epidermal Growth Factor Receptor Expression Status in Patients With Non-Small Cell Lung Cancer
Xuemei Huang,Yingli Sun,Mingyu Tan,Weiling Ma,Pan Gao,Lin Qi,Jinjuan Lu,Yuling Yang,Kun Wang,Wufei Chen,Liang Jin,Kaiming Kuang,Shaofeng Duan,Ming Li
DOI: https://doi.org/10.3389/fonc.2022.772770
IF: 4.7
2022-02-02
Frontiers in Oncology
Abstract:Objectives EGFR testing is a mandatory step before targeted therapy for non-small cell lung cancer patients. Combining some quantifiable features to establish a predictive model of EGFR expression status, break the limitations of tissue biopsy. Materials and Methods We retrospectively analyzed 1074 patients of non-small cell lung cancer with complete reports of EGFR gene testing. Then manually segmented VOI, captured the clinicopathological features, analyzed traditional radiology features, and extracted radiomic, and deep learning features. The cases were randomly divided into training and test set. We carried out feature screening; then applied the light GBM algorithm, Resnet-101 algorithm, logistic regression to develop sole models, and fused models to predict EGFR mutation conditions. The efficiency of models was evaluated by ROC and PRC curves. Results We successfully established Model clinical , Model radiomic , Model CNN (based on clinical-radiology, radiomic and deep learning features respectively), Model radiomic+clinical (combining clinical-radiology and radiomic features), and Model CNN+radiomic+clinical (combining clinical-radiology, radiomic, and deep learning features). Among the prediction models, Model CNN+radiomic+clinical showed the highest performance, followed by Model CNN , and then Model radiomic+clinical . All three models were able to accurately predict EGFR mutation with AUC values of 0.751, 0.738, and 0.684, respectively. There was no significant difference in the AUC values between Model CNN+radiomic+clinical and Model CNN . Further analysis showed that Model CNN+radiomic+clinical effectively improved the efficacy of Model radiomic+clinical and showed better efficacy than Model CNN . The inclusion of clinical-radiology features did not effectively improve the efficacy of Model radiomic . Conclusions Either deep learning or radiomic signature-based models can provide a fairly accurate non-invasive prediction of EGFR expression status. The model combined both features effectively enhanced the performance of radiomic models and provided marginal enhancement to deep learning models. Collectively, fusion models offer a novel and more reliable way of providing the efficacy of currently developed prediction models, and have far-reaching potential for the optimization of noninvasive EGFR mutation status prediction methods.
oncology