作者: Yang Song , Weidong Cai , Yun Zhou , Dagan Feng
DOI: 10.1007/978-3-642-33415-3_10
关键词:
摘要: Automatic detection of lung tumors and abnormal lymph nodes are useful in assisting cancer staging. This paper presents a novel method, by first identifying all abnormalities, then differentiating between based on their degree overlap with the field mediastinum. Regression-based appearance model graph-based structure labeling designed to estimate actual mediastinum from pathology-affected thoracic images adaptively. The proposed method is simple, effective generalizable, can be potentially applicable other medical imaging domains as well. Promising results demonstrated our evaluations clinical PET-CT data sets patients.