BACKGROUND: Noninvasively detecting epidermal growth factor receptor (EGFR) mutation status in lung adenocarcinoma patients before targeted therapy remains a challenge. This study aimed to develop a 3-dimensional (3D) convolutional neural network (CNN)-based deep learning model to predict EGFR mutation status using computed tomography (CT) images. METHODS: We retrospectively collected 660 patients from 2 large medical centers. The patients were divided into training (n=528) and external test (n=132) sets according to hospital source. The CNN model was trained in a supervised end-to-end manner, and its performance was evaluated using an external test set. To compare the performance of the CNN model, we constructed 1 clinical and 3 radiomics models. Furthermore, we constructed a comprehensive model combining the highest-performing radiomics and CNN models. The receiver operating characteristic (ROC) curves were used as primary measures of performance for each model. Delong test was used to compare performance differences between different models. RESULTS: Compared with the clinical [training set, area under the curve (AUC) =69.6%, 95% confidence interval (CI), 0.661-0.732; test set, AUC =68.4%, 95% CI, 0.609-0.752] and the highest-performing radiomics models (training set, AUC =84.3%, 95% CI, 0.812-0.873; test set, AUC =72.4%, 95% CI, 0.653-0.794) models, the CNN model (training set, AUC =94.3%, 95% CI, 0.920-0.961; test set, AUC =94.7%, 95% CI, 0.894-0.978) had significantly better predictive performance for predicting EGFR mutation status. In addition, compared with the comprehensive model (training set, AUC =95.7%, 95% CI, 0.942-0.971; test set, AUC =87.4%, 95% CI, 0.820-0.924), the CNN model had better stability. CONCLUSIONS: The CNN model has excellent performance in non-invasively predicting EGFR mutation status in patients with lung adenocarcinoma and is expected to become an auxiliary tool for clinicians.

Background

To explore the value of dual-energy spectral CT in distinguishing solitary pulmonary tuberculosis (SP-TB) from solitary lung adenocarcinoma (S-LUAD).

Methods

A total of 246 patients confirmed SP-TB (n = 86) or S-LUAD (n = 160) were retrospectively included. Spectral CT parameters include CT_40keV value, CT_70keV value, iodine concentration (IC), water concentration (WC), effective atomic number (Zeff), and spectral curve slope (λ_70keV). Data were measured during the arterial phase (AP) and venous phase (VP). Chi-square test was used to compare categorical variables, Wilcoxon rank-sum test was used to compare continuous variables, and a two-sample t-test was used to compare spectral CT parameters. ROC curves were used to calculate diagnostic efficiency.

Results

There were significant differences in spectral CT quantitative parameters (including CT_40keV value [all P< 0.001] , CT_70keV value [all P< 0.001], λ_70keV [P< 0.001, and P = 0.027], Zeff [P =0.015, and P = 0.001], and IC [P =0.002, and P = 0.028]) between the two groups during the AP and VP. However, WC (P = 0.930, and P = 0.823) was not statistically different between the two groups. The ROC curve analysis showed that the AUC in the AP and VP was 90.9% (95% CI, 0.873-0.945) and 83.4% (95% CI, 0.780-0.887), respectively. The highest diagnostic performance (AUC, 97.6%; 95% CI, 0.961-0.991) was achieved when all spectral CT parameters were combined with clinical variables.

Conclusion

Dual-energy spectral CT has a significant value in distinguishing SP-TB from S-LUAD.

Background

This study aimed to noninvasively predict the mutation status of epidermal growth factor receptor (EGFR) molecular subtype in lung adenocarcinoma based on CT radiomics features.

Methods

In total, 728 patients with lung adenocarcinoma were included, and divided into three groups according to EGFR mutation subtypes. 1727 radiomics features were extracted from the three-dimensional images of each patient. Wilcoxon test, least absolute shrinkage and selection operator regression, and multiple logistic regression were used for feature selection. ROC curve was used to evaluate the predictive performance of the model. Nomogram was constructed by combining radiomics features and clinical risk factors. Calibration curve was used to evaluate the goodness of fit of the model. Decision curve analysis was used to evaluate the clinical applicability of the model.

Results

There were three, two, and one clinical factor and fourteen, thirteen, and four radiomics features, respectively, which were significantly related to each EGFR molecular subtype. Compared with the clinical and radiomics models, the combined model had the highest predictive performance in predicting EGFR molecular subtypes [Del-19 mutation vs. wild-type, AUC=0.838 (95% CI, 0.799-0.877); L858R mutation vs. wild-type, AUC=0.855 (95% CI, 0.817-0.894); and Del-19 mutation vs. L858R mutation, AUC=0.906 (95% CI, 0.869-0.943), respectively], and it has a stable performance in the validation set [AUC was 0.813 (95% CI, 0.740-0.886), 0.852 (95% CI, 0.790-0.913), and 0.875 (95% CI, 0.781-0.929), respectively].

Conclusion

Our combined model showed good performance in predicting EGFR molecular subtypes in patients with lung adenocarcinoma. This model can be applied to patients with lung adenocarcinoma.