A multicenter validation of an active contour-based left ventricular analysis technique

Abstract

Quantitative analysis of functional cardiac magnetic resonance (MR) images has been limited by the lack of well‐validated, semiautomatic, methods for rapid analysis. We describe the evaluation of a DICOM‐compatible PC‐based parallel‐processing tool, for cardiac magnetic resonance analysis (CAMRA), which supports semiautomatic image mensuration using an active contour model‐based algorithm. The CAMRA software was used to analyze data from 12 patients in a multicenter acquisition and analysis trial to compare semiautomatic contour detection with manual planimetry of the left ventricular endocardium from short‐axis, breath‐held, cine gradient‐echo images. There was excellent agreement between the manual and semiautomatic measurements of global left ventricular function, with no significant (P = 0.32) difference in the determination of ejection fraction (−0.9 ± 3.1% [mean difference ± 1 standard deviation]). There was no significant interobserver difference in the semiautomatically measured ejection fraction. Additionally, a single observer completed the analysis on data from 30 patients and found no significant (P = 0.05) difference in the determination of ejection fraction (−1.3 ± 3.5% [mean difference ± 1 standard deviation]). The CAMRA software demonstrates the capability for the reproducible evaluation of global left ventricular function in cardiac patients, with adequate interobserver reproducibility for use in multicenter trials. J. Magn. Reson. Imaging 2000;12:232–239. © 2000 Wiley‐Liss, Inc.

Comparative evaluation of active contour model extensions for automated cardiac MR image segmentation by regional error assessment

“…However the tendency of semi-automated methods to underestimate endocardial volumes and to overestimate epicardial volumes, with no significant difference in the final calculation of EF, was confirmed. The marked differences between manual and semi-automated methods in the present paper can be explained by the absence of manual correction after automated segmentation [13,27], as well as no image exclusion due to automatic segmentation failure or unsatisfactory image quality [12]. The second reason is the method of propagation of the contour between slices.…”

“…In order to reduce the variability of the measurements and time constraints, a large number of automatic and semi-automatic procedures, based on various image processing approaches, have been proposed in the last few years. These procedures include thresholding and shape extraction [5], region growing [6,7], graph searching [8,9], deformable models [10][11][12][13], loop B-spline curve fitting [14], and fuzzy clustering [15,16]. Among these methods, the active contour model, or snake, introduced by Kass [17] and representing a special case of the general multidimensional deformable model theory, has attracted most of the attention to date and has been extensively studied and used with promising results.…”

“…In comparison to others studies [12,13,27], the segmentation results were less accurate. However the tendency of semi-automated methods to underestimate endocardial volumes and to overestimate epicardial volumes, with no significant difference in the final calculation of EF, was confirmed.…”

“…However clinical validation of these algorithms has been described in studies comparing automated and manual segmentations [12,13,27] with end-diastolic volume (EDV), end-systolic volume (ESV), and ejection fraction (EF) measurements. Graves et al [13] showed that EDV and ESV were underestimated with semi-automatic methods, but there was no significant difference in the final calculation of EF. Van der Geest et al [27] described smaller endocardial contours and larger epicardial contours with a semi-automated detection algorithm.…”

An automated myocardial segmentation in cardiac MRI

“…Moreover the important variation of size and shape of a pathological myocardium represents a challenge of the model construction. Deformable models have also been widely used for segmenting cardiac images [7], [8], [9], [10], [11]. Active contour models have been quite successful for segmenting the myocardium boundaries using CMR images.…”