Abstract for Oral Presentation Automatic determination of LV volume in MSCT using 3D active shape models (ASMs) Hans C. van Assen, Boudewijn P.F. Lelieveldt, Mikhail G. Danilouchkine, Rob J. van der Geest & Johan H.C. Reiber Division of Image Processing, Department of Radiology, Leiden University Medical Center, Leiden, The Netherlands Cardiac CT provides new opportunities to assess ventricular function in combination with coronary morphology. However, manual analysis of ventricular function is prohibitively labor intensive. We developed an automatic contour detection technique based on 3D active shape model (ASM): an adaptive contour detection method that can learn from manually drawn expert contours. A Cardiovascular Imaging 2002 – NASCI Symposium 2002 493 statistical model was constructed using contours from 53 patients, resulting in 33 eigenvariations, which describe 99% of the shape variations in the training set. The model-matching algorithm was evaluated using CT data from six patients, from which the contours generated by the 3D ASM were compared to expert drawn contours. Comparisons were only carried out for slices in which automatically generated contours were present. Endocardial volume was calculated from the areas enclosed by the contours using Simpson’s rule. The error in endocardial volume ranged from )2.2 to +7.7% with respect to the expert determined volumes. From the calculated volumes a regression formula was derived: y ¼ 0.95x + 2.8 (R ¼ 0.99), with y being the volume resulting from the ASM contours and x, the volume resulting from the expert drawn contours. Based on these results we conclude that the 3D ASM is a promising tool for left ventricular volume determination from MSCT.
|Journal||The International Journal of Cardiovascular Imaging|
|Publication status||Published - 2002|