To evaluate the potential of computed tomography (CT) radiomics, based on high-resolution large matrix target reconstruction images, in predicting the invasiveness of lung adenocarcinoma in pure ground-glass nodules (pGGNs) with a diameter ≤ 1.5 cm. The clinical and imaging data of 297 patients with pGGNs, confirmed by pathology, were collected between March 2021 and June 2024. Pathological diagnoses included atypical adenomatous hyperplasia (AAH), adenocarcinoma in situ (AIS), minimally invasive adenocarcinoma (MIA) and invasive adenocarcinoma (IAC). The patients were divided into non-invasive (AAH and AIS) and invasive (MIA and IAC) groups based on pathology. Radiomics features were extracted using ITK-SNAP software, and a predictive model was built using Python 3.9.7, with feature selection based on least absolute shrinkage and selection operator regression. Receiver operating characteristic analysis, area under the curve (AUC), sensitivity, specificity and clinical decision curve analysis were used to assess model performance. Multivariate logistic regression revealed that the maximum lesion diameter, median CT value and solid component ratio were significant predictors of invasiveness (P < 0.05). The CT radiomics model achieved AUC values of 0.861 (95% confidence interval CI 0.811–0.912) in the training set and 0.790 (95% CI 0.687–0.892) in the validation set. A combined model integrating clinical and radiomics features showed improved predictive performance, with an AUC of 0.861 (95% CI 0.809–0.913) in the training set and 0.810 (95% CI 0.709–0.912) in the validation set. The combined model based on CT radiomics and clinical imaging showed good performance in predicting the invasiveness of small pGGNs and may assist in clinical decision-making regarding follow-up management, surgery timing and treatment strategies. Further validation in prospective, multicentre studies is needed to verify these findings and assess generalisability to broader populations.
Lv et al. (Wed,) studied this question.