Xiaojie Huang

Contour tracking in echocardiographic sequences via sparse representation and dictionary learning.

This paper presents a dynamical appearance model based on sparse representation and dictionary learning for tracking both endocardial and epicardial contours of the left ventricle in echocardiographic sequences. Instead of learning offline spatiotemporal priors from databases, we exploit the inherent spatiotemporal coherence of individual data to constraint cardiac contour estimation. The contour tracker is initialized with a manual tracing of the first frame. It employs multiscale sparse representation of local image appearance and learns online multiscale appearance dictionaries in a boosting framework as the image sequence is segmented frame-by-frame sequentially. The weights of multiscale appearance dictionaries are optimized automatically. Our region-based level set segmentation integrates a spectrum of complementary multilevel information including intensity, multiscale local appearance, and dynamical shape prediction. The approach is validated on twenty-six 4D canine echocardiographic images acquired from both healthy and post-infarct canines. The segmentation results agree well with expert manual tracings. The ejection fraction estimates also show good agreement with manual results. Advantages of our approach are demonstrated by comparisons with a conventional pure intensity model, a registration-based contour tracker, and a state-of-the-art database-dependent offline dynamical shape model. We also demonstrate the feasibility of clinical application by applying the method to four 4D human data sets.

Active Contours with Group Similarity

Active contours are widely used in image segmentation. To cope with missing or misleading features in images, researchers have introduced various ways to model the prior of shapes and use the prior to constrain active contours. However, the shape prior is usually learnt from a large set of annotated data, which is not always accessible in practice. Moreover, it is often doubted that the existing shapes in the training set will be sufficient to model the new instance in the testing image. In this paper, we propose to use the group similarity of object shapes in multiple images as a prior to aid segmentation, which can be interpreted as an unsupervised approach of shape prior modeling. We show that the rank of the matrix consisting of multiple shapes is a good measure of the group similarity of the shapes, and the nuclear norm minimization is a simple and effective way to impose the proposed constraint on existing active contour models. Moreover, we develop a fast algorithm to solve the proposed model by using the accelerated proximal method. Experiments using echocardiographic image sequences acquired from acute canine experiments demonstrate that the proposed method can consistently improve the performance of active contour models and increase the robustness against image defects such as missing boundaries.

Radial Basis Functions for Combining Shape and Speckle Tracking in 4D Echocardiography

Quantitative analysis of left ventricular deformation can provide valuable information about the extent of disease as well as the efficacy of treatment. In this work, we develop an adaptive multi-level compactly supported radial basis approach for deformation analysis in 3D+time echocardiography. Our method combines displacement information from shape tracking of myocardial boundaries (derived from B-mode data) with mid-wall displacements from radio-frequency-based ultrasound speckle tracking. We evaluate our methods on open-chest canines (N=8) and show that our combined approach is better correlated to magnetic resonance tagging-derived strains than either individual method. We also are able to identify regions of myocardial infarction (confirmed by postmortem analysis) using radial strain values obtained with our approach.

A Dynamical Appearance Model Based on Multiscale Sparse Representation: Segmentation of the Left Ventricle from 4D Echocardiography

The spatio-temporal coherence in data plays an important role in echocardiographic segmentation. While learning offline dynamical priors from databases has received considerable attention, these priors may not be suitable for post-infarct patients and children with congenital heart disease. This paper presents a dynamical appearance model (DAM) driven by individual inherent data coherence. It employs multi-scale sparse representation of local appearance, learns online multiscale appearance dictionaries as the image sequence is segmented sequentially, and integrates a spectrum of complementary multiscale appearance information including intensity, multiscale local appearance, and dynamical shape predictions. It overcomes the limitations of database-driven statistical models and applies to a broader range of subjects. Results on 26 4D canine echocardiographic images acquired from both healthy and post-infarct subjects show that our method significantly improves segmentation accuracy and robustness compared to a conventional intensity model and our previous single-scale sparse representation method.

Segmentation of left ventricles from echocardiographic sequences via sparse appearance representation

Sparse representation has proven to be a powerful mathematical framework for studying high-dimensional data and uncovering its structures. Some recent research has shown its promise in discriminating image patterns. This paper presents an approach employing sparse appearance representation for segmenting left ventricular endocardial and epicardial boundaries from 2D echocardiographic sequences. It leverages the inherent spatio-temporal coherence of tissue/blood appearance over the sequence by modeling the different appearance of blood and tissues with different appearance dictionaries and updating the dictionaries in a boosting framework as the frames are segmented sequentially. The appearance of each frame is predicted in the form of appearance dictionaries based on the appearance observed in the preceding frames. The dictionaries discriminate image patterns by reconstructing them in the process of sparse coding resulting in an appearance discriminant that we incorporate into a region-based level set segmentation process. We illustrate the advantages of our approach by comparing it to manual tracings and an intensity-prior-based level set method. Experimental results on 34 2D canine echocardiographic sequences show that sparse appearance representation significantly outperforms intensity in terms of reliability and accuracy of segmentation.

Segmentation of 4D Echocardiography Using Stochastic Online Dictionary Learning

Dictionary learning has been shown to be effective in exploiting spatiotemporal coherence for echocardiographic segmentation. To overcome the limitations of previous methods, we present a stochastic online dictionary learning approach for segmenting left ventricular borders from 4D echocardiography. It is based on stochastic approximations and processes a mini-batch of samples at a time, which results in lower memory consumption and lower computational cost than classical batch algorithms. In contrast to the previous methods, where dictionaries and their weights are optimized only on the most recently segmented frame, our stochastic online learning procedure optimizes the dictionaries and the corresponding weights by aggregating all the past information while adapting them to the dynamically changing data. The rate of updating the past information is controlled and varied according to the appearance scale to seek a balance between old and new information. Results on 26 4D echocardiographic images show the proposed method is more accurate, more robust, and faster than the previous batch algorithm.

A combined shape tracking and speckle tracking approach for 4D deformation analysis in echocardiography

Quantitative analysis of left ventricular function provides valuable information about overall heart health. Echocardiography is non-invasive method for imaging the heart that provides high temporal resolution allowing for imaging cardiac motion. Shape tracking and speckle tracking are two methods that have previously been used to track cardiac deformation in ultrasound. These two methods provide complementary displacement values. Shape tracking gives accurate boundary displacements, while speckle tracking provides accurate displacements across the myocardium. We combine these two methods using adaptive multilevel radial basis functions on 4D echocardiography data. Radial strains were calculated for N=5 open chest canines 6 weeks post-coronary artery occlusion. Radial strain values for three myocardial segments across three short axis slices showed good correlation to corresponding MR tagged data.

4-D echocardiography assessment of local myocardial strain using 3-D speckle tracking combined with shape tracking

In 4-D echocardiography (4DE), displacement estimates obtained solely from multi-dimensional speckle tracking can exhibit large variances and peak hopping, making it challenging to accurately calculate myocardial strains. 3-D phase-sensitive speckle tracking can produce sensitive estimates along the axial direction, but typically provides poorer estimates in orthogonal directions and at tissue boundaries. Shape tracking provides complimentary information, as it effectively tracks myocardial boundaries and does not depend on beam orientation. We propose a method combining 3-D speckle tracking with 3-D shape tracking using a quality-based radial basis function approach. Echocardiographic data (3D+t) were acquired in an open chest canine model at six weeks following surgical coronary occlusion using a commercial 2-D phased array, on which 3-D phase-sensitive speckle tracking and 3-D shape tracking were performed. An adaptive, multi-level radial basis function method was used to combine information from the two tracking methods, utilizing confidence metrics to weight the contribution of each estimate to generate a dense 3-D displacement field throughout the myocardium. A multi-level approach was used to capture smaller scales of motion in regions of fine deformation variation and high tracking confidence. The 3-D combined approach produced displacement estimates with greatly reduced variance and peak hopping compared to 3-D speckle tracking alone. Lower radial strains were observed in the myocardial infarct region, corresponding to reduced local contractility. Strong correlations were observed for both radial and circumferential strains between the combined method and estimates from magnetic resonance (MR) tagging studies.