Jigar Bharat Patel MD, Presenter: Nothing to Disclose

Thomas Flukinger, Abstract Co-Author: Research support, Riverain Medical

Ryan Moffitt, Abstract Co-Author: Employee, TeraRecon, Inc

Eliot Lawrence Siegel MD, Abstract Co-Author: Nothing to Disclose

Khan Mohammad Siddiqui MD, Abstract Co-Author: Partner, iVirtuoso, Inc, Baltimore, MD Medical Advisory Board, General Electric Company, Barrington, IL Research Consultant, Mercury Computer Systems, Inc, Chelmsford, MA

Julia Iddings Flukinger MD, Abstract Co-Author: Spouse, employee, Riverain Medical, Dayton, OH

The purpose of this study was to assess the accuracy and utility of an automated anatomic mapping system in the identification and annotation of cardiac and coronary anatomy on routine coronary CT angiography (CCTA).

A prototype 3D commercial application server was used to perform automated anatomy detection, mapping, and labeling of cardiac and coronary structures on 30 different CCTA acquisitions at varying phases of cardiac cycle in 3 patients (mean age, 52 y). Each examination was performed on a 64-detector CT scanner with retrospective ECG gating with the following parameters: slice thickness, 1 mm; slice increment, 0.5 mm; reconstruction interval, 1 mm; 500 mAs; 140 kVp. A nonionic contrast medium (150 mL)was infused intravenously at a rate of 5 mL/sec followed by normal saline (30 mL) at 5 mL/sec. Automated bolus tracking was employed when the descending aorta reached 150 HU. Anatomic structures mapped and labeled included the aorta, cardiac chambers, valves and coronary vessels. Manual validation of anatomic mapping was performed and the data was analyzed to assess labeling accuracy.

Ten separate cardiac structures were labeled in 25 interpretable cardiac phases. Correct labeling for the ascending aorta was 84%, aortic root 80%, cardiac apex 80%, left ventricle 52%, mitral valve 40%, left main coronary artery (CA) 52%, mid left anterior descending CA 28%, mid left circumflex CA 52%, proximal right CA 56%, and mid right CA 16%. The percentages of these 10 structures identified on each phase were 85% for the 0% phase, 60% for 10% phase, 40% for 20% phase, 20% for 30% phase, 33% for 40% phase, 35% for 50% phase, 43% for 60% phase, 70% for 70% phase, 95% for 80% phase, and 75% for the 90% phase.

Automated anatomic labeling software can be used for the identification of cardiac structures on reconstructed cardiac CT images. Identification accuracy is optimal in the cardiac phases near end-diastole (70%-90% phases).

Automated identification of cardiac anatomy has the potential to cut down on labor-intensive cardiac postprocessing and provide region-specific processing based on individual anatomic structure.

Patel, J, Flukinger, T, Moffitt, R, Siegel, E, Siddiqui, K, Flukinger, J, Automated Mapping of Cardiac and Coronary Anatomy for Coronary CT Angiography.  Radiological Society of North America 2007 Scientific Assembly and Annual Meeting, November 25 - November 30, 2007 ,Chicago IL. http://archive.rsna.org/2007/5016165.html