Artificial intelligence-assisted quantitative CT parameters in predicting the degree of risk of solitary pulmonary nodules

IntroductionArtificial intelligence (AI) shows promise for evaluating solitary pulmonary nodules (SPNs) on computed tomography (CT). Accurately determining cancer invasiveness can guide treatment. We aimed to investigate quantitative CT parameters for invasiveness prediction.MethodsPatients with stage 0-IB NSCLC after surgical resection were retrospectively analysed. Preoperative CTs were evaluated with specialized software for nodule segmentation and CT quantification. Pathology was the reference for invasiveness. Univariate and multivariate logistic regression assessed predictors of high-risk SPN.ResultsThree hundred and fifty-five SPN were included. On multivariate analysis, CT value mean and nodule type (ground glass opacity vs. solid) were independent predictors of high-risk SPN. The area under the curve (AUC) was 0.811 for identifying high-risk nodules.ConclusionsQuantitative CT measures and nodule type correlated with invasiveness. Software-based CT assessment shows potential for noninvasive prediction to guide extent of resection. Further prospective validation is needed, including comparison with benign nodules.