Affiliation:
1. Guangzhou Women and Children's Medical Center, Guangzhou Medical University
2. Xiangtan Central Hospital
3. Affiliated Hospital of Guilin Medical University
Abstract
Abstract
Background
Radiomic applications for differentiating clinical stage IA solitary pulmonary nodule (SPN)-type invasive mucinous adenocarcinoma (IMA) from SPN-presenting lung adenocarcinoma (LADC) are lacking. Therefore, this study aimed to develop and validate predictive models for the preoperative differentiation between SPN-IMA and invasive non-mucinous LADC using computed tomography (CT) radiological and radiomic features.
Methods
In this bicentric study, we collected 507 SPNs, of which 42 were diagnosed as IMA and 465 as invasive non-mucinous LADC. The patients were randomly divided into training and test sets at a ratio of 7:3. The minimal redundancy maximal relevance filter was used to extract radiomic features, and the least absolute shrinkage and selection operator regression was used to screen these features and calculate the individualized radiomic score (rad score). We constructed a prediction nomogram that integrated radiomics and CT radiological features by applying multivariate logistic regression. Diagnostic capabilities were assessed by comparing the receiver operating characteristic and area under the curve (AUC) values.
Results
The combined model achieved AUC values of 0.789 and 0.798 for the training and test sets, respectively, surpassing those of the radiomics model in both the training (p = 0.038) and test (p = 0.021) sets. Moreover, the combined model performed better than the clinical model in the training (p = 0.017) and test (p = 0.025) sets. We transformed this combined model into a nomogram that accurately quantifies the risk of IMA and demonstrates exceptional discrimination and calibration.
Conclusions
The combined nomogram, incorporating radiomics and CT radiological features, is potentially valuable for the preoperative differentiation between clinical stage IA SPN-type IMA and invasive non-mucinous LADC.
Publisher
Research Square Platform LLC