CT radiomics based model for differentiating malignant and benign small (≤20mm) solid pulmonary nodules

Abstract Background At present, computed tomography (CT) radiomics-based models capable of evaluating small (≤ 20 mm) solid pulmonary nodules (SPNs) are lacking. Accordingly, the present study sought to develop a CT radiomics-based model capable of differentiating between benign and malignant small SPNs. Methods Between January 2019 and November 2021, this study enrolled consecutive patients presenting with small SPNs, randomly assigning these individuals to training and testing cohorts at an 8:2 ratio. CT images were processed to extract radiomics features, with a radiomics scoring model being developed based on the features selected in the training group through univariate and multivariate logistic regression analyses. The testing cohort was then used to validate the developed predictive model. Results In total, this study included 210 patients in the training (n = 168) and testing (n = 42) cohorts. Radiomics scores were ultimately calculated based on 9 selected CT radiomics features. Traditional CT and clinical risk factors associated with malignancy in SPNs included lobulation (P < 0.001), spiculation (P < 0.001), and a larger diameter (P < 0.001). The developed CT radiomics scoring model consisted of the following formula: X = -6.773 + 12.0705×radiomics score + 2.5313×lobulation + 3.1761×spiculation + 0.3253×diameter. The CT radiomics-based model, CT radiomics score, and clinicoradiological score were associated with area under the curve (AUC) values of 0.957, 0.945, and 0.853, respectively, in the training cohort, while the testing cohort exhibited corresponding AUC values of 0.943, 0.916, and 0.816. Conclusions The CT radiomics-based model designed in the present study offers valuable diagnostic accuracy when employed to distinguish between benign and malignant SPNs.