Application of computed tomography-based radiomics combined with clinical factors in the diagnosis of malignant degree of lung adenocarcinoma.

Background: As an emerging technology, radiomics is being widely used in the diagnosis of early lung cancer due to its excellent diagnostic performance. However, there is a lack of studies that apply radiomics to the diagnosis of malignancy of lung adenocarcinoma. Thus, we used computed tomography (CT)-based radiomics to construct a model for the diagnosis of high-risk lung adenocarcinoma. Methods: Data of 170 patients who underwent surgical treatment at the First Affiliated Hospital of Soochow University and had a maximum nodule diameter ≤2 cm on preoperative CT images between January 2020 and December 2021 were retrospectively analyzed. All enrolled patients were randomly divided into experimental and validation groups according to the ratio of 7:3. The diagnosis of lung adenocarcinoma was based on postoperative pathological results. The region of interest was delineated on preoperative CT images, and the radiomics features were extracted. The least absolute shrinkage and selection operator (LASSO) was used to screen the radiomics features thus obtaining the radiomics score (Radscore), which was the basis of the radiomics model. Based on the multivariate regression analysis, independent predictors were screened from the clinical baseline data and imaging features thus constructing clinical model. Multivariate logistic regression was used to combine independent predictors and the Radscore to form a comprehensive nomogram. The diagnostic performance of constructed models was evaluated based on receiver operating characteristic (ROC) curves and decision curve analysis (DCA). Results: The sensitivity and specificity of the clinical model based on consolidation-to-tumor ratio (CTR), lobulated signs and vascular anomaly signs was 70.0% and 76.7% in the validation group. The radiomics model [area under the curve (AUC) 0.926; 95% confidence interval (CI): 0.857-0.995] and the comprehensive model (AUC 0.922; 95% CI: 0.851-0.992) performed better than clinical model (AUC 0.839; 95% CI: 0.720-0.958) in the validation group. The sensitivity and specificity of the comprehensive model was 85.0% and 80.0% in the validation group. DCA of radiomics model and comprehensive model suggested they have better net survival benefit than clinical model. Conclusions: Compared with clinical model, radiomics model and comprehensive model had better diagnostic performance in distinguishing malignant degree of lung adenocarcinoma.