Bi-temporal 3D active appearance models with applications to unsupervised ejection fraction estimation

Abstract

Rapid and unsupervised quantitative analysis is of utmost importance to ensure clinical acceptance of many examinations using cardiac magnetic resonance imaging (MRI). We present a framework that aims at fulfilling these goals for the application of left ventricular ejection fraction estimation in four-dimensional MRI. The theoretical foundation of our work is the generative two-dimensional Active Appearance Models by Cootes et al., here extended to bi-temporal, three-dimensional models. Further issues treated include correction of respiratory induced slice displacements, systole detection, and a texture model pruning strategy. Cross-validation carried out on clinical-quality scans of twelve volunteers indicates that ejection fraction and cardiac blood pool volumes can be estimated automatically and rapidly with accuracy on par with typical inter-observer variability.

Automated frame‐by‐frame endocardial border detection from cardiac magnetic resonance images for quantitative assessment of left ventricular function: Validation and clinical feasibility

“…Different techniques for endocardial contour detection from cardiac MR images have been described in the literature (14), including several reports on semiautomated and automated techniques (15)(16)(17)(18)(19)(20)(21)(22)(23), many of which were suitable for analysis of ED and ES phases only (15)(16)(17) . Limitations of these approaches include a large number of manually segmented images required to build a model database (18 -20) or to train the model (21)(22), no continuous temporal segmentation (18), and computational complexity (23).…”

A Novel Explicit 2D+t Cyclic Shape Model Applied to Echocardiography

“…in Active Shape Models (ASMs) [4] or Active Appearance Models (AAMs) [5,6]. Some medical imaging modalities, especially echocardiography, depend strongly on temporal information, and implicit time extensions have been proposed to 2D and 3D PCA models [7,8]. Such models are implicit because instead of adding a time variable, they are built from the concatenation of shape vectors…”

Echocardiographic Image Sequence Compression Based on Spatial Active Appearance Model

“…Having the distribution points established, the AAM will be enabled to adjust itself to a diversity of biological factors like the phase of ECG and breathing. A detailed description of the manifestation of these phenomena and the model adaptation is given in [18]. Our algorithm acts similarly, but it treats the cardiac cycle differently: not only systolic and diastolic phases are distinguished, but also a QRS complex clustering is performed to give different treatment to normal and ventricular cardiac cycles.…”

“…Several papers have already reported the usage of spatial AAM [13,18]. The present work has the following contributions: (1) we developed a heart reconstruction algorithm including time-dependent wall boundaries, to estimate the image variances, that allow a better compression rate than conventional methods at a fixed image quality; (2) reported techniques classify ultrasound images only as belonging to systolic or diastolic interval.…”