70O - Radiomic-based quantitative CT analysis to predict the invasiveness of pure ground-glass lung adenocarcinoma

Abstract

Background: Lung cancer is till one of the leading causes of cancer-related deaths and lung adenocarcinoma is the most common type. Compared with invasive adenocarcinoma, adenocarcinoma in situ (AIS) and minimally invasive adenocarcinoma (MIA) are considered as indolent lung adenocarcinoma with good prognosis. However, AIS, MIA and even some early-stage invasive adenocarcinoma can be shown as pure ground-glass nodules on computed tomorgraphy images, which is quite difficult for clinicians to make a precise diagnosis and a suitable treatment plan. Thus, we aim to investigate the performance of radiomic-based quantitative analysis on CT images in identifying invasiveness of lung adenocarcinoma manifesting as pure ground-glass nodules. Methods: 275 lung adenocarcinoma cases with 322 pure ground-glass nodules from January 2015 to October 2017 were enrolled in this retrospective study. All lesions were resected surgically and confirmed pathologically. Clinical data like age, gender, smoking status of all cases were collected from digital medical records. Radiomic feature extraction was performed using Python with semi-automatically segmented tumor regions on CT scans which was contoured with an in-house developed plugin for 3D-Slicer. The predictive performance of the prediction models was evaluated through the receiver operating characteristic curve (ROC). Results: Among 322 nodules, 48(15%) were Adenocarcinoma in situ (AIS), 102(32%) were minimally invasive adenocarcinoma (MIA) and 172(53%) were invasive adenocarcinoma. All nodules were divided into training and validation cohort randomly with a ratio of 2:1 to establish prediction models. The values of the area under the curve were 0.716 (95%CI:0.600 0.832) and 0.827 (95%CI:0.729 0.925) with the diagnostic accuracy of 59.3% and 69.1% for radiomic and combined models, respectively. Conclusions: Radiomic model built via quantitative CT analysis can help to identify the invasiveness of lung adenocarcinoma represented as pure ground-glass nodules. Combining this model with clinical features can significantly improve its prediction performance. Legal entity responsible for the study: Fangyi Xv. Funding: Has not received any funding. Disclosure: All authors have declared no conflicts of interest.