Objective to develop machine learning models based on radiomics features extracted from the primary tumor, peritumoral region, and habitat region to predict lymph node metastasis status in patients with invasive pulmonary adenocarcinoma, and to evaluate the predictive performance and generalizability of different radiomic features. Methods A total of 1 263 patients with invasive pulmonary adenocarcinoma who underwent surgery at the Department of Thoracic Surgery, Jiangsu Provincial People’s Hospital from 2016 to 2019 were enrolled. The habitat regions on CT images were delineated by applying K-means clustering (with an average cluster number of 2) based on grayscale values. The peritumoral region was defined as a uniform 3 mm expansion around the primary tumor, which was segmented using an automatic V-net model combined with manual correction. Radiomic features were then extracted from these predefined regions respectively, and stacking machine learning models were constructed. The model performance was evaluated on the training, testing, and internal validation cohorts using metrics including the area under the receiver operating characteristic curve (AUC), F1 score, recall, and precision. Results After excluding patients who did not meet the inclusion criteria, 651 patients were finally included in the study, consisting of 181 males and 287 females, with a mean age of 58.39±11.23 years (range 29-78). Although the habitat radiomics-based model did not show the best performance in the training set, it demonstrated outstanding results in the internal validation set, achieving an AUC of 0.95 [95% CI (0.87, 1.00)], an F1 score of 0.85, and a precision-recall area under the curve (PR-AUC) of 0.89, outperforming models based on the primary tumor and peritumoral regions.ConclusionThe model constructed based on habitat radiomics exhibited superior performance in the internal validation set, suggesting it may possess better generalizability and clinical utility in predicting lymph node metastasis status in pulmonary adenocarcinoma.