- Poster presentation
- Open Access
Clinical evaluation of automatic whole-heart and coronary-artery segmentation
© Breeuwer et al; licensee BioMed Central Ltd. 2009
- Published: 28 January 2009
- Manual Correction
- Deformable Model
- Segmentation Quality
- Correction Effort
- Fair Quality
With Steady-State Free-Precession (SSFP) MRI the heart and surrounding arteries can be imaged. The resulting image data can be used to inspect the patient-specific cardiac anatomy and to quantify aspects such as left-ventricular volume and wall mass. The data can furthermore be used to detect stenosis in the proximal and middle coronary-artery segments . To simplify the visualization, quantification and stenosis detection, we have developed an almost automatic whole-heart and coronary-artery segmentation method.
We describe the clinical evaluation of the performance of our cardiac MR whole-heart and coronary-artery segmentation method.
Our segmentation method was first developed for Computed Tomography  and thereafter adapted to MR . The method fully automatically segments the ventricles, atria, and part of the aorta and pulmonary vessels by means of a shape-constrained deformable model, which has been constructed from a large set of representative trainings data. Manual corrections of the segmentation are possible by single mouse clicks on the 3D segmentation surface visualization. The points where the coronary-arteries branch off the aorta (ostia) are also automatically detected. The user has to manually indicate the coronary-artery end points in the image data, these arteries are thereafter automatically tracked . Total time needed for automatic segmentation/tracking is less than 2 min per data set (Dell 670 PC, dual-processor 3 GHz, 3 Gbyte memory).
Our method was evaluated by an experienced clinical cardiologist (third author) on whole-heart MR acquisitions from 50 patients with coronary-artery disease or heart failure (Philips Achieva 1.5 T, typically 150 slices, TE = 2.14 ms, TR = 4.27 ms, flip angle = 86 degree, pixel spacing = 0.7 mm, slice distance = 0.8 mm). The cardiologist visually judged the segmentation quality before and after manual correction (categories: poor, fair, average, good, excellent). The time needed to perform segmentation corrections was recorded and the cardiologist judged the required correction effort (categories: none, small, medium, large).
For 47 out of 50 investigated cases (94%), our automatic whole-heart and coronary-artery segmentation method in combination with small or medium correction effort is judged to result in good or excellent segmentation quality.
- Yong Kim W, et al: New Engl J Med. 2001, 345: 1863-1869. 10.1056/NEJMoa010866.View ArticleGoogle Scholar
- Ecabert O, et al: Proc SPIE. 2007, 6512: 65120G.Google Scholar
- Peters J, et al: MICCAI. 2007, 402-410. Part II, LNCS 4792Google Scholar
- Sonnemans J: J Cardiovas Magn Reson. 2008, 10 (Suppl 1): Proceedings SCMR 2008.Google Scholar
This article is published under license to BioMed Central Ltd.