Joanne Hoffman BS, BA, Abstract Co-Author: Nothing to Disclose

Jiamin Liu PhD, Abstract Co-Author: Nothing to Disclose

Evrim Bengi Turkbey MD, Abstract Co-Author: Nothing to Disclose

Ronald M. Summers MD, PhD, Presenter: Royalties, iCAD, Inc Research funded, iCAD, Inc Stockholder, Johnson & Johnson Grant, Viatronix, Inc

Station-labeling of mediastinal lymph nodes is typically performed manually by qualitative visual assessment in clinical radiology practice, which is very time consuming. In this work, we present a method of automatically recognizing the mediastinal lymph node stations by multi-atlas label fusion organ segmentation in contrast-enhanced thoracic CT scans.

83 patients with contrasted-enhanced CT scans were analyzed. These patients were randomly selected by searching for the terms “lymphadenopathy” and “mediastinal” in radiology reports. Each patient had at least one lymph node measuring over 10 mm in short axis in the mediastinum. The data set included 329 lymph nodes with sizes greater than 10 mm (15.6±4.9mm) found on the scans. Following the International Association for the Study of Lung Cancer (IASLC) definitions, 84 hilar and lobar (stations 10, 11, 12), 72 inferior mediastinal (stations 7, 8, 9), 123 superior mediastinal (stations 2L, 2R, 3A, 3P, 4R, 4L), 28 aortic (stations 5 and 6), and 22 other (para-cardiac, para-aortic, para-esophageal) lymph nodes were labeled and verified by a radiologist as the reference standard. The station label for each lymph node was automatically determined by fully-automated computer software that identifies 8 organs using multi-atlas label fusion and the relative position of the node to these organs in order to assign the IASLC station location.  

Our method achieved high accuracy in labeling the correct station for lymph nodes in these stations: superior mediastinal (82.9%), hilar and lobar (90.5%), inferior mediastinal (91.7%) and other (95.5%). Low accuracy was found in the aortic stations (39.2%) which accounted for only 28/329 lymph nodes. Of all lymph nodes, 83.9% (276/329) were correctly labeled.  

This method accurately and fully-automatically determines mediastinal lymph node station location on thoracic CT. The method can be used on automatically generated detections and radiologist designated lymph nodes.  

The accurate and precise identification of the lymph node stations on computed tomography (CT) images is important for proper staging, prognosis, and treatment assessment in patients with cancer.

Hoffman, J, Liu, J, Turkbey, E, Summers, R, Automatic Localization of IASLC-defined Mediastinal Lymph Node Stations on CT Scans Using Multi-atlas Multi-organ Segmentation.  Radiological Society of North America 2014 Scientific Assembly and Annual Meeting, - ,Chicago IL. http://archive.rsna.org/2014/14007908.html