Combining CT radiomics and radiological features to predict pathological grade in stage I lung adenocarcinoma.

Aims: While previous studies have explored the use of radiomics or radiological features alone, this study uniquely integrates both feature types within a unified predictive model based on the latest IASLC grading system, thereby enhancing pathological grading accuracy in early-stage invasive lung adenocarcinoma. This study aimed to evaluate the potential of combining CT radiomics with traditional radiological features to non-invasively predict the pathological grade of stage I invasive pulmonary adenocarcinoma according to the International Association for the Study of Lung Cancer (IASLC) new grading system.

Methods: A retrospective study was conducted on 240 patients. Radiological features were assessed, and radiomics texture features were selected using mRMR and LASSO. A combined predictive model was constructed using random forest regression, and its diagnostic performance was evaluated using ROC analysis.

Results: The CT radiological feature model achieved AUCs of 0.848 in the training set and 0.832 in the validation set. The texture feature model yielded AUCs of 0.850 in the training set and 0.845 in the validation set. The combined predictive model demonstrated superior diagnostic performance with AUCs of 0.902 in the training set and 0.880 in the validation set. The combined model's specificity also exceeded that of the individual models, with specificities of 90.5% and 93.3% in the training and validation sets, respectively.

Conclusions: This model improved diagnostic accuracy and demonstrated strong potential for clinical application, advocating for its broader adoption in clinical practice to improve personalized treatment strategies in lung cancer care.