Detection of pathological areas and estimation of viability parameters in late-enhancement cardiac MRI
- PDF / 662,681 Bytes
- 2 Pages / 610 x 792 pts Page_size
- 20 Downloads / 165 Views
BioMed Central
Open Access
Poster presentation
Detection of pathological areas and estimation of viability parameters in late-enhancement cardiac MRI Khaoula Elagouni1, Cybele Ciofolo-Veit*1, Benoît Mory1, Gilion Hautvast2 and Marcel Breeuwer2 Address: 1Philips Medical Systems Research Paris, Suresnes Cedex, France and 2Philips Healthcare, Best, Netherlands * Corresponding author
from 13th Annual SCMR Scientific Sessions Phoenix, AZ, USA. 21-24 January 2010 Published: 21 January 2010 Journal of Cardiovascular Magnetic Resonance 2010, 12(Suppl 1):P167
doi:10.1186/1532-429X-12-S1-P167
Abstracts of the 13th Annual SCMR Scientific Sessions - 2010
Meeting abstracts - A single PDF containing all abstracts in this Supplement is available here. http://www.biomedcentral.com/content/files/pdf/1532-429X-11-S1-infoThis abstract is available from: http://jcmr-online.com/content/12/S1/P167 © 2010 Elagouni et al; licensee BioMed Central Ltd.
Introduction
Results
Viability assessment is essential for therapy planning following a myocardial infarction. In particular, the proportion of viable myocardium is a major factor in determining whether a patient may benefit from revascularization [1]. By means of late-enhancement cardiac MR imaging (LECMR), it is now possible to visualize normal and non-viable areas with high spatial resolution.
The method was tested on 11 short-axis LECMR volumes (100 slices) acquired in a multi-center study between 2004-2007 (Philips Intera scanner 1.5 T, M FFE sequence, TE = 1.1 ms, TR = 3.8 ms, flip angle = 15°). All slices are 256 × 256, spaced with 10 mm and with pixel size around 1.5 mm. An experienced operator used the interactive thresholding procedure available in the Philips ViewForum 4.3 Cardiac Package to locate pathological areas. To automatically detect these areas we used: (1) our method (see results in Figure 1), (2) thresholding at average normal myocardium value plus two standard deviations (SD) and (3) thresholding at average normal myocardium value plus three SD. Following segmentation, we quantified the volumetric fraction of pathologic areas and the distance between the damaged region and the epicardium. We compared the results obtained from automatic segmentation to the results obtained from manual segmentation (Figure 2). Our method gives the smallest error with respect to manual detections (detected tissue error = 3.5 +/- 5.6% of heart volume, distance error = 12 +/- 11% of myocardium thickness) and always avoided false detections contrary to thresholding approaches.
Purpose We propose a novel fully automatic method to detect and quantify pathological areas in late-enhancement cardiac MR images.
Methods The first step consists to analyse myocardium intensity with a specifically tailored Expectation-Maximization algorithm [2]. Estimated distributions of both normal and abnormal tissues are then exploited to generate a fuzzy map indicating, for every pixel, the membership degree to the abnormal tissue class. As false positives remain, we design a fast version of
Data Loading...
