Abstract
BACKGROUND: Accurately assessing the invasiveness of lung adenocarcinoma with part-solid pulmonary nodules is pivotal for the treatment. Recent studies have highlighted the crucial role of adipose in the growth and invasion of tumors. This study aimed to develop and validate a model by integrating radiomic features from intrathoracic adipose tissue (IAT) with clinical factors to classify lung adenocarcinoma invasiveness. METHODS: 608 patients with lung adenocarcinomas were enrolled in three different centers. Radiomic features were extracted from nodules and IAT in computed tomography images. Multivariable logistic regression analysis was used to develop a nomogram model. The nomogram's performance was assessed utilizing accuracy, discrimination, and clinical benefits. The net reclassification index (NRI) and integrated discrimination improvement (IDI) were applied to evaluate changes in model performance after the addition of IAT features. RESULTS: The nomogram, incorporating nodule and IAT signature with nodular diameter, was constructed in the training dataset. The nomogram demonstrated area under the receiver operating curve values of 0.927 (95% CI: 0.872,0.983), 0.945 (95% CI: 0.902,0.988), and 0.917 (95% CI: 0.871,0.962) in the internal validation and two external validation cohorts, respectively. The nomogram showed good calibration (Hosmer-Lemeshow test, P>0.05), and adding the IAT signatures did improve nomogram performance in all cohorts (all NRI and IDI >0, P<0.05). Decision curve and stratification analysis revealed that the nomogram was clinically useful and had potential generalization ability. CONCLUSIONS: The nomogram, consisting of nodule signature, IAT signature, and clinical factors, could individualize the classification of invasive adenocarcinomas with part-solid pulmonary nodules.