Correlation between high-resolution computed tomography lung nodule characteristics and EGFR mutation in lung adenocarcinomas

Three-Dimensional Convolutional Neural Network-Based Prediction of Epidermal Growth Factor Receptor Expression Status in Patients With Non-Small Cell Lung Cancer

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.

Predictors of Invasive Adenocarcinomas among Pure Ground-Glass Nodules Less Than 2 cm in Diameter

Simple Summary

Benign lesions, atypical adenomatous hyperplasia, and malignancies such as adenocarcinoma in situ, minimally invasive adenocarcinoma, and invasive adenocarcinoma may feature pure ground-glass nodules on chest CT images, and the prognosis of patients with invasive adenocarcinoma is worse than others. The early detection and adequate management of invasive adenocarcinoma is crucial, but the pathology diagnosis of small nodules is difficult to obtain without surgery. Our study aimed to analyze the CT characteristics of pure ground-glass nodules <2 cm for the identification of invasive adenocarcinomas. A total of 181 nodules in 171 patients were enrolled. The larger size, lobulation, and air cavity were significantly more common in invasive adenocarcinoma. The air cavity is the significant predictor in multivariate analysis. In conclusion, the possibility of invasive adenocarcinoma is higher in a pure ground-glass nodules when it is associated with a larger size, lobulation, and air cavity.

Abstract

Benign lesions, atypical adenomatous hyperplasia (AAH), and malignancies such as adenocarcinoma in situ (AIS), minimally invasive adenocarcinoma (MIA), and invasive adenocarcinoma (IA) may feature a pure ground-glass nodule (pGGN) on a thin-slide computed tomography (CT) image. According to the World Health Organization (WHO) classification for lung cancer, the prognosis of patients with IA is worse than those with AIS and MIA. It is relatively risky to perform a core needle biopsy of a pGGN less than 2 cm to obtain a reliable pathological diagnosis. The early and adequate management of patients with IA may provide a favorable prognosis. This study aimed to disclose suggestive signs of CT to accurately predict IA among the pGGNs. A total of 181 pGGNs of less than 2 cm, in 171 patients who had preoperative CT-guided localization for surgical excision of a lung nodule between December 2013 and August 2019, were enrolled. All had CT images of 0.625 mm slice thickness during CT-guided intervention to confirm that the nodules were purely ground glass. The clinical data, CT images, and pathological reports of those 171 patients were reviewed. The CT findings of pGGNs including the location, the maximal diameter in the long axis (size-L), the maximal short axis diameter perpendicular to the size-L (size-S), and the mean value of long and short axis diameters (size-M), internal content, shape, interface, margin, lobulation, spiculation, air cavity, vessel relationship, and pleural retraction were recorded and analyzed. The final pathological diagnoses of the 181 pGGNs comprised 29 benign nodules, 14 AAHs, 25 AISs, 55 MIAs, and 58 IAs. Statistical analysis showed that there were significant differences among the aforementioned five groups with respect to size-L, size-S, and size-M (p = 0.029, 0.043, 0.025, respectively). In the univariate analysis, there were significant differences between the invasive adenocarcinomas and the non-invasive adenocarcinomas with respect to the size-L, size-S, size-M, lobulation, and air cavity (p = 0.009, 0.016, 0.008, 0.031, 0.004, respectively) between the invasive adenocarcinomas and the non-invasive adenocarcinomas. The receiver operating characteristic (ROC) curve of size for discriminating invasive adenocarcinoma also revealed similar area under curve (AUC) values among size-L (0.620), size-S (0.614), and size-M (0.623). The cut-off value of 7 mm in size-M had a sensitivity of 50.0% and a specificity of 76.4% for detecting IAs. In the multivariate analysis, the presence of air cavity was a significant predictor of IA (p = 0.042). In conclusion, the possibility of IA is higher in a pGGN when it is associated with a larger size, lobulation, and air cavity. The air cavity is the significant predictor of IA.