Deciphering the intratumoral histologic heterogeneity of lung adenocarcinoma using radiomics

This study utilized radiomics to predict high-risk histologic patterns in lung adenocarcinoma (LUAD) and analyze their association with occult lymph node metastasis and recurrence rate. Data from 777 patients (Institution A: 528, Institution B: 249) across two medical institutions were collected, and high-resolution preoperative CT images were analyzed to extract radiomic features. Using data from Institution A, a logistic regression-based model was developed to predict micropapillary and solid patterns, and external validation was conducted with Institution B data. A composite model incorporating six CT-based radiomic features and clinical variables outperformed single models (AUC = 0.84–0.91). Lung nodule segmentation was performed using AVIEW software from Coreline Soft, followed by expert modification. This study demonstrates the potential of preoperative, non-invasive prediction of high-risk histologic patterns in LUAD.