Catálogo de publicaciones - libros

Automating the analysis of left ventricular (LV) wall motion can improve objective prediction of coronary artery disease. A new method for classifying LV wall motion using shape models with localized variations was developed for this purpose. These sparse shape models were built from four-chamber and two-chamber echocardiographic sequences using principal component analysis and orthomax rotations. The resulting shape parameters were then used to classify wall-motion abnormalities of LV segments. Compared with the shape model before rotation, higher classification correctness was achieved using significantly less shape parameters. The local variations exhibited by these shape parameters correlated reasonably with the location of the segments.

This paper presents an enhanced version of our previous algorithm for point-wise tracking and analysis of cardiac motion based on 3D active mesh model. In the present software, a new 3D active surface model based on curve evolution techniques and level sets is used for automatizing the segmentation of endocardial boundary in the left ventricle in the first phase of cardiac cycle. Furthermore, cardiac muscle anisotropy is modeled and included in the tracking algorithm. Additionally, the tracking algorithm is improved in order to track the left ventriclular wall instead of left ventricular cavity in the previous version. Finally, a quantitative analysis of myocardial strain is performed using the motion estimation obtained by the tracking software. Experiments were performed on cardiac MRI images for tracking the left ventricle myocardium. The results of evaluation on a set of Gradient-Echo images reported in this paper clearly demonstrate the effectiveness of our algorithm for extracting motion parameters.

Tagged Magnetic Resonance Imaging (tagged MRI or tMRI) provides a means of directly and noninvasively displaying the motion of the myocardium. Reconstruction of the motion field is needed for quantitative analysis of important clinical information, e.g., the myocardial strain. In this paper, we present a two-step method for this task. First, we use a Gabor filter bank to generate a corresponding phase map of tMRI images. Second, deformable models are initialized at the discontinuities in the wrapped phase map, and are deformed under the influence of the image gradient to track the motion of tags. Unlike previous approaches, a Robust Point Matching (RPM) module has been integrated into the model evolution to avoid false tracking results caused by 1) through-plane motion, and 2) small tag spacing. The method has been tested on a numeric phantom, as well as in vivo heart data. The experimental results show that the new method has a good performance on both synthetic and real data, and has the potential to be used in clinical applications.

The objective of this study was to validate a deformable image registration technique, termed Hyperelastic Warping, for left ventricular strain measurement during the systole using cine-gated nontagged MRI with strains measured from tagged MRI. Tagged and non-tagged cine images were obtained on a 1.5 T Siemens Avanto clinical scanner with a TrueFISP imaging sequence. The Hyperelastic Warping solution was evolved using a series of non-tagged images in 10 phases from end-diastole to end-systole. The solution may be considered as ten separate Warping problems with multiple Templates and Targets. At each stage, an active contraction was initially applied to the FE model, and then Warping penalty forces were utilized to generate the final registration. Warping results for circumferential strain were correlated (R^2 =0.59) with results obtain from tagged MR images analyzed with a HARP algorithm. Results for fiber stretch, LV twist, and transmural strain distribution were similar to values in the literature. Hyperelastic Warping represents a novel approach for quantifying 3-D regional strains within the myocardium with a high resolution.

In this paper we extend the Frangi filter[1] to recognize edges and do not enhance them. We give a theoretical framework for optimal scale selection and choice of the free parameters . We discuss discretization details concerning especially the discrete kernel used for building the scale-space and the choice of discrete scales. We present several experiments on phantom data to objectively and quantitatively compare and judge the filters. Experiments on real coronary angiograms enhance the improvement reached by the integration of the edge indicator.

SENC is a new technique for imaging tissue deformation, such as the strain of cardiac tissue due to contraction. SENC strain quantifications are limited to one direction, the through-plane direction. However, this is sufficient to image circumferential and longitudinal strain in the long- and short-axis views, respectively. The factors that affect the accuracy of SENC strain mesurements are the slice profile and the signal-to-noise ratio (SNR). In this work, these factors are analyzed in order to optimize the SENC method for strain quantifications.

We present a system that reconstructs the 3D motion of the left ventricle (LV) for a full cardiac cycle using a deformable model built from tagged MR images. Two sets of cues are drawn from tagged MRI. The intersections of the three tagging planes, and the intersections of the LV boundary and the tagging planes, are interpolated onto the mesh vertices. We implement a deformable model to track the LV motion, utilizing Finite Element Methods (FEM) to keep the general shape and topology of the LV. This volumetric deformable model speeds up the FEM and facilitates the medical analysis. The LV motion reconstruction provides information for further analysis of cardiac mechanisms.

A computer aided reconstruction and motion analysis method of mitral annulus is presented in this paper. To begin with, the boundary points on mitral annulus are marked by doctors interactively. Since these points are not distributed uniformly and sequentially, secondly, it is necessary to re-arrange these points into a set of series points on a contour, the saddle-shaped mitral annulus. Thirdly, in order to analyze 3D mitral annulus motion, the mitral annulus is modeled by 3D non-uniform rational B-spline (NURBS). Fourthly, the dynamic parameters of the mitral annulus throughout the cardiac cycle are computed in a 3D Cartesian coordinate system. The experiments prove that the dynamic mitral annulus reconstruction and analysis program using computer aided method is provided a possible and convenient tool to diagnose and analyze the malfunction of mitral annulus.

This paper presents a volumetric cardiac analysis and movement reconstruction algorithm from echocardiographic image sequences and electrocardiography (ECG) records. The method consists of two-dimensional (2-D) echocardiogram transformation, shape detection, heart wall movement identification, volumetric analysis and 4-D model construction. Although the semi-periodic behavior of the ECG and the breath caused heart rate variance disturbs spatial and temporal reconstruction, the presented algorithm is able to overcome these problems in most cases for normal and ventricular beats. The obtained model provides a tool to investigate volumetric variance of the heart and the phenomenon of normal and abnormal heart beating that makes possible to explore continuously the heart’s inner structure.

Magnetic resonance diffusion tensor imaging (DTI) has emerged as a convenient and reliable alternative to conventional histology for characterizing the fiber structure of the myocardium. The acquisition of full data for different diffusion directions for a large number of slices often takes a long time and results in trade-offs in the number of slices and signal to noise ratios. We propose a constrained reconstruction technique based on a regularization framework to jointly reconstruct sparse sets of cardiac DTI data. Constraints on spatial variation and directional variation were used in the reconstruction. The method was tested on sparse data undersampled in both rectilinear and (simulated) radial fashions and compared to reconstructions from full data. The method provided reasonable reconstructions with half of the data for rectilinear undersampling and similar quality images with a quarter of the data if radial undersampling was used.