Automated visualization tool aids coronary CT angiography
Global left ventricular (LV) function is perhaps the strongest determinant of heart failure and death due to myocardial infarction. Traditionally, LV function studies conducted with CT have been performed with 2D planimetry of short-axis CT image reformations. New advances in automated advanced visualization 3D technology can produce comparable results to this methodology, according to a group of German researchers, without the shortcomings of the 2D approach.
“This 2D approach has the limitation that numerous user interactions are time-consuming and introduce systematic error in definition of the short-axis plane and determination of the endocardial contours and most basal and apical sections,” wrote the authors of a multi-institution study published in this month’s American Journal of Roentgenology.
The researchers utilized a dual-source CT system (Somatom Definition, Siemens Medical Solutions) to conduct angiography studies on 50 consecutively registered patients with known or suspected coronary artery disease. The exam data was then processed for axial and short-axis CT image reformations at the scanner workstation (Syngo CT-2007A, VA 10A, Siemens).
The axial images were transferred to an Aquarius (TeraRecon) workstation, where an assessment of the LV function was performed with a dedicated software package for cardiac function analysis (TVA version 3.5.2.1, TeraRecon) with threshold-based region-growing 3D segmentation of the LV cavity, according to the scientists. The software automatically segmented the contrast-filled LV lumen for all 20 cardiac phases.
The short-axis CT image reformations were evaluated on a Leonardo (Siemens) workstation, using a standard cardiac CT software package (Syngo Argus, Siemens).
“The end-diastolic and end-systolic phases were visually determined as the phases with the largest and smallest, respectively, ventricular cavities,” the authors wrote. Endocardial borders were contoured semi-automatically and checked visually for accuracy on all CT images with a discernible LV cavity. The most basal slice was defined as the image closest to the mitral valve annulus showing LV myocardium in at least 50 percent of its perimeter. The most apical image was the last image with a visible LV lumen. Papillary muscles were included as part of the LV cavity.”
The LV functional analysis of the automated and semi-automated approach was conducted by a clinician with eight years of experience in cardiac radiology.
“Mean analysis time with the 2D short-axis approach was 248 ± 29 seconds,” the researchers reported. “Threshold-based 3D segmentation analysis was performed within the significantly shorter mean time of 172 ± 20 seconds (p < 0.05).”
The scientists reported that there was excellent correlation between the 2D and 3D approach for LV end-diastolic and end-systolic volumes.
“Automated threshold-based 3D segmentation allows accurate and reproducible estimation of LV volumes and function from dual-source CT coronary angiographic data sets in close correlation with results obtained with the established 2D short-axis approach,” they wrote.
“This 2D approach has the limitation that numerous user interactions are time-consuming and introduce systematic error in definition of the short-axis plane and determination of the endocardial contours and most basal and apical sections,” wrote the authors of a multi-institution study published in this month’s American Journal of Roentgenology.
The researchers utilized a dual-source CT system (Somatom Definition, Siemens Medical Solutions) to conduct angiography studies on 50 consecutively registered patients with known or suspected coronary artery disease. The exam data was then processed for axial and short-axis CT image reformations at the scanner workstation (Syngo CT-2007A, VA 10A, Siemens).
The axial images were transferred to an Aquarius (TeraRecon) workstation, where an assessment of the LV function was performed with a dedicated software package for cardiac function analysis (TVA version 3.5.2.1, TeraRecon) with threshold-based region-growing 3D segmentation of the LV cavity, according to the scientists. The software automatically segmented the contrast-filled LV lumen for all 20 cardiac phases.
The short-axis CT image reformations were evaluated on a Leonardo (Siemens) workstation, using a standard cardiac CT software package (Syngo Argus, Siemens).
“The end-diastolic and end-systolic phases were visually determined as the phases with the largest and smallest, respectively, ventricular cavities,” the authors wrote. Endocardial borders were contoured semi-automatically and checked visually for accuracy on all CT images with a discernible LV cavity. The most basal slice was defined as the image closest to the mitral valve annulus showing LV myocardium in at least 50 percent of its perimeter. The most apical image was the last image with a visible LV lumen. Papillary muscles were included as part of the LV cavity.”
The LV functional analysis of the automated and semi-automated approach was conducted by a clinician with eight years of experience in cardiac radiology.
“Mean analysis time with the 2D short-axis approach was 248 ± 29 seconds,” the researchers reported. “Threshold-based 3D segmentation analysis was performed within the significantly shorter mean time of 172 ± 20 seconds (p < 0.05).”
 |
| A 68-year-old man with three-vessel coronary artery disease and inferolateral myocardial infarction. A–F, Diastolic (A–C) and systolic (D–F) CT image reformations show reduced diastolic wall thickness and absence of systolic wall thickening of inferior and inferolateral walls of left ventricle. Left ventricular volumes and ejection fraction determined with 3D threshold segmentation algorithm (A, B, D, E) and 2D short axis-based planimetry (Simpson method) (C, F) show comparable results. Image and caption by permission of the American Roentgen Ray Society. |
The scientists reported that there was excellent correlation between the 2D and 3D approach for LV end-diastolic and end-systolic volumes.
“Automated threshold-based 3D segmentation allows accurate and reproducible estimation of LV volumes and function from dual-source CT coronary angiographic data sets in close correlation with results obtained with the established 2D short-axis approach,” they wrote.