CT-derived Body Composition Associated with Pulmonary Nodule Malignancy and Growth

ObjectiveThis study investigated the association between body composition and pulmonary nodule malignancy and growth. MethodsA dataset of subjects with indeterminate pulmonary nodules (IPNs) was created from an internal (n=216) and external (n=162) cohort. Five different body tissues were automatically segmented and quantified from baseline and follow-up chest low-dose computed tomography (LDCT) scans using artificial intelligence (AI) algorithms. Logistic Regression (LR) analyses, t-tests, and Person correlation analyses were performed to study the association between body tissues and nodule malignancy, as well as nodule changes such as density, size, and shape. Gender differences were investigated. The area under the receiver operating characteristic curve (ROC-AUC) was used to assess classifier performance. Average feature importance was evaluated using several machine learning models. Causal relationships were analyzed and visualized using a novel directed graph method. ResultsUnivariate analysis revealed a significant association between Skeletal muscle density and nodule malignancy in both genders (p<0.001). The multivariate model based on body composition yielded AUCs of 0.77 (95% CI: 0.71 - 0.84) and 0.63 (95% CI: 0.54 - 0.72) on the internal and external datasets, respectively. The composite model based on body composition and nodule features yielded AUCs of 0.87 (95% CI: 0.82 - 0.91) and 0.62 (95% CI: 0.53 - 0.72) on the internal and external datasets, respectively. Skeletal muscle and intermuscular adipose tissue features were highly ranked among tissue features, with skeletal muscle density retaining its highest rank even after adjusting for clinical and nodule features. The causal graph identified two nodule features and skeletal muscle density as directly linked to nodule malignancy. Skeletal muscle density and intramuscular adipose tissue density were identified as nodule growth indicators in both genders. ConclusionsBody composition can serve as a potential biomarker for assessing nodule malignancy and evaluating nodule growth in both genders. Summary StatementWe found that body composition were critical indicators for discriminating malignant nodules from benign ones and for evaluating the nodule growth in both males and females. Key ResultsO_LIUnivariate analysis revealed a significant association between body composition and nodule malignancy in both males and females. Multivariate analysis further demonstrated the predictive ability of body composition features. C_LIO_LIFeature importance analysis and causal graph analysis identified skeletal muscle density as one of the leading features associated with nodule malignancy. C_LIO_LISkeletal muscle density and intramuscular adipose tissue density were identified as nodule growth indicators in both males and females. C_LI