GAN-generated target reconstruction CT radiomics for prediction of the novel IASLC grading of pulmonary adenocarcinoma

作者：Qinling Jiang, Wei Deng, Jiang X, Hongbiao Sun, Xiang Wang, Xianpan Pan, Zirui Cao, Gang Xiang, Ya Wen, Q Chen, Yuxin Cheng, Tianyi Xing, Lei Chen, Yao Xiao · 发表于：Journal of Thoracic Disease · 年份：2026 · DOI：10.21037/jtd-2026-0717 · 研究领域：Radiomics and Machine Learning in Medical Imaging、Lung Cancer Diagnosis and Treatment、MRI in cancer diagnosis

Background: The novel International Association for the Study of Lung Cancer (IASLC) pathologic grading system provides a better stratification of prognosis. This study aims to develop a radiomics model based on the generative-adversarial-network (GAN)-generated target reconstruction computed tomography (GTRCT) for the preoperative prediction of IASLC grading and to compare that with the normal-resolution computed tomography (NRCT). Methods: This retrospective study included patients pathologically diagnosed with invasive pulmonary adenocarcinoma (IPA) from July 2018 to December 2021 from Changzheng Hospital and The Affiliated Traditional Chinese Medicine Hospital of Southwest Medical University. Among them, 122 patients from Hospital 2 served as a separate external validation set. By extracting image features and using least absolute shrinkage and selection operator to screen the features, radiomics models were constructed to predict the novel IASLC grading of IPA, and the model performance of GTRCT and NRCT was compared. The diagnostic performance of the models was evaluated using receiver operating characteristic curves (ROC), accuracy, sensitivity, and specificity. In addition, calibration curves and decision curve analysis (DCA) were applied to validate two different models. Results: This retrospective study included 396 patients (mean age: 60.41±10.35 years; male: 184) from the two hospitals. Radiomics models based on GTRCT and NRCT both demonstrated good predictive per...