Skip to Main Content
Positron emission tomography - computed tomography (PETCT) is now accepted as the best imaging technique to accurately stage lung cancer. The consistent and accurate interpretation of PET-CT images, however, is not a trivial task. We propose a content-based image retrieval system for retrieving similar cases from an imaging database as a reference dataset to aid the physicians in PET-CT scan interpretation. Problematic areas in diagnosis are the abnormal FDG uptake in the parenchymal lung tumor and in the regional nodes in the pulmonary hilar regions and the mediastinum. The primary tumor and the nodal disease are detected from the scans of thorax with learning-based techniques and a voting method for 3D object localization. Similar cases are then retrieved based on the similarity measure between the feature vectors of the cases. Our preliminary evaluation with clinical data from lung cancer patients suggests our approach is accurate with high retrieval precision.