Catálogo de publicaciones - libros

In this paper we present a novel approach to the problem of spatio-temporal alignment of cardiac MR image sequences. This novel method has the ability to correct spatial misalignment caused by global acquisition and local shape differences as well as temporal misalignment caused by differences in the length of the cardiac cycles or by differences in the motion patterns of the hearts. In contrast to our previous approach [1], the algorithm optimizes the spatial and temporal transformation components separately, thus significantly speeding up the registration process. To achieve this we have developed a novel approach for the calculation of the temporal registration which does not require the spatial alignment of the image sequences. The method was evaluated using fifteen cardiac MR image sequences from healthy volunteers and the results were compared to the previously presented method. The results indicate that the performance of the method is similar to the previously presented method [1] while the the computational complexity has been significantly reduced.

This paper presents a comparison of the motion fields computed from TrueFISP untagged and SPAMM tagged magnetic resonance (MR) images using a 4D nonrigid registration algorithm that we have developed for cardiac motion tracking [3]. Our results, which were obtained from a group of 7 normal volunteers, indicate that although there is a good correlation between the motion fields computed from the tagged and untagged MR images, some of the twisting motion is not captured in the motion fields derived from the TrueFISP MR images.

Matrix-phased array transducers for real-time three-dimensional ultrasound enable fast, non-invasive visualization of cardiac ventricles. Segmentation of 3D ultrasound is typically performed at end diastole and end systole with challenges for automation of the process and propagation of segmentation in time. In this context, given the position of the endocardial surface at certain instants in the cardiac cycle, automated tracking of the surface over the remaining time frames could reduce the workload of cardiologists and optimize analysis of volume ultrasound data. In this paper, we applied optical flow to track the endocardial surface between frames of reference, segmented via manual tracing or manual editing of the output from a deformable model. To evaluate optical-flow tracking of the endocardium, quantitative comparison of ventricular geometry and dynamic cardiac function are reported on two open-chest dog data sets and a clinical data set. Results showed excellent agreement between optical flow tracking and segmented surfaces at reference frames, suggesting that optical flow can provide dynamic “interpolation” of a segmented endocardial surface.

Magnetic resonance tagging technique measures the deformation of the heart wall by overlying darker tag lines onto the brighter myocardium and tracking their motion during the heart cycle. In this paper, we propose a new spline-based methodology for constructing a dense cardiac displacement map based on the tag tracking result. In this new approach, the deformed tags are tracked using a Gabor filter-based technique and smoothed using implicit splines. Then we measure the displacement in the myocardium of both ventricles using a new spline interpolation model. This model uses rough segmentation results to set up break points along tag tracking spline so that the local myocardium deformation will not be influenced by the tag information in the blood or the deformation in other parts of the myocardium. The displacements in x- and y-directions are calculated separately and are combined later to form the final displacement map. This method accepts either a tag grid or separate horizontal and vertical tag lines as its input by adjusting the offsets of images taken at different breath hold. The method can compute dense displacement maps of the myocardium for time phases during systole and diastole. The approach has been quantatively validated on phantom images and been tested on more than 20 sets of in-vivo heart data.

Multislice Computed Tomography (MSCT) scanners offers new perspectives for cardiac kinetics evaluation with 3D time image sequences of high contrast and spatio-temporal resolutions. A new method is proposed for cardiac motion extraction in Multislice CT. Based on a 3D surface-volume matching process, it provides the detection of the heart left cavities along the acquired sequence and the estimation of their 3D surface velocity fields. A 3D segmentation step and surface reconstruction process are first applied on only one image of the sequence to obtain a 3D mesh representation for one time. A Markov Random Field model is defined to find best correspondences between 3D mesh nodes at time and voxels in the next volume at + 1 time. A simulated annealing is used to perform a global optimization of the correspondences. First results obtained on simulated and real data show the good behaviour of this method.

In the context of motion estimation of the heart and thoracic structures from tomographic imaging, we investigated two free form deformations (FFD) based non linear registration methods as motion estimators. Standard and cylindrical FFD (CFFD) methods are evaluated in 2D, both on simulated and in vivo cardiac and thoracic images. Results tend to show that CFFD based method achieves the same accuracy with less parameters. However, the fast convergence of this model is hamped by a higher computing time with a straightforward implantation.

The interpretation of cine-loops and parametric images to assess regional wall motion in echocardiography requires to acquire an expertise, which is based on training. To overcome the training phase for the interpretation of new parametric images, a quantification based on profiles in the parametric images was attempted. The classification of motion was performed on a training set including 362 segments and tested on a second database including 238 segments. The consensual visual interpretation of two-dimensional sequences by two experienced readers were used as the ”gold standard”. Mono- and multi-parametric classification approaches were undertaken. Results show an accuracy of 74% for training and 68% for test in case of mono-parametric approach. They are 80% and 67% in case of multi-parametric approach. Moreover, the evaluation protocol enables to understand the limitations of this approach. The in-depth study shows that a large part of false-positive segments are apical segments. This suggests that taking into account the segment location could improve the performances.