Lung cancer is the leading cause of cancer-related death worldwide.The 5-year survival rate of patients with stage I–IV lung adenocarcinoma is merely 15%. According to the classification of lung adenocarcinoma proposed by International Association for the Study of Lung Cancer, the American Thoracic Society and the European Respiratory Society (IASLC/ATS/ERS) in 2011, lung adenocarcinoma is divided into three categories: pre-invasive lesions (atypical adenomatous hyperplasia [AAH] and adenocarcinoma in situ [AIS]), minimally invasive adenocarcinoma (MIA) and invasive adenocarcinoma (IA). AIS and MIA are appropriate indications for sublobar resection (wedge resection or segmentectomy), which could reserve more lung function without compromising survival as well as reduced operation-related complications and mortalities.Thus, discriminating between IA and non-IA (including AIS and MIA) is crucial for predicting the prognosis of lung adenocarcinoma and selecting the proper surgical treatment plan.

Our research investigated and developed a new deep learning scheme to classify between IA and non-IA by using residual learning architecture. The flowchart of our scheme is illustrated in Figure 1.

• Authour: Jing Gong
• E-mail: gongjing1990@163.com
• Citation: Gong J, Liu J, Hao W, et al. A deep residual learning network for predicting lung adenocarcinoma manifesting as ground-glass nodule on CT images[J]. European Radiology, 2019: 1-9.(DOI: 10.1007/s00330-019-06533-w)