Elsa D. Angelini

Dynamic Cardiac Information From Optical Flow Using Four Dimensional Ultrasound

Quantitative analysis of cardiac motion is of great clinical interest in assessing ventricular function. Real-time 3-D (RT3D) ultrasound transducers provide valuable three-dimensional information, from which quantitative measures of cardiac function can be extracted. Such analysis requires segmentation and visual tracking of the left ventricular endocardial border. We present results based on correlation of four-dimensional optical flow motion for temporal tracking of ventricular borders in three dimensional ultrasound data. A displacement field is computed from the optical flow output, and a framework for the computation of dynamic cardiac information is introduced. The method was applied to a clinical data set from a heart transplant patient and dynamic measurements agreed with physiological knowledge as well as experimental results

Tracking of LV endocardial surface on real-time three-dimensional ultrasound with optical flow

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.

Review of Myocardial Motion Estimation Methods from Optical Flow Tracking on Ultrasound Data

Quantitative analysis of cardiac motion is of great clinical interest in assessment of ventricular function. Ultrasound imaging, especially matrix transducers acquiring real-time three dimensional data, provide valuable information, from which quantitative measures of cardiac function can be extracted via optical flow computation. Such analysis requires tracking of the image brightness patterns with underlying assumptions of visual persistency. We present a review of myocardial motion analysis on cardiac ultrasound, based on optical-flow computation, with phantom and clinical evaluations for segmental wall assessment and motion features quantification

VALIDATION OF OPTICAL-FLOW FOR QUANTIFICATION OF MYOCARDIAL DEFORMATIONS ON SIMULATED RT3D ULTRASOUND

Quantitative analysis of cardiac motion is of great clinical interest in assessing ventricular function. Real-time 3-D (RT3D) ultrasound transducers provide valuable four-dimensional information, from which quantitative measures of cardiac function can be extracted. Previously, we presented a method based on four-dimensional optical flow motion estimation for anatomical tracking of myocardium in RT3D ultrasound, from which myocardial displacement fields and dynamic cardiac metrics were computed. In this paper, in order to quantitatively validate our method, we build a truly 3D mathematical phantom of cardiac tissue and blood. Distinguished from previous studies, our work further decomposes tissue impedance into cell kernels and processes all functions in 3D. Instead of simply modeling the myocardium, a quasi-LV phantom is built including myocardium and blood. Also all ultrasound probe parameters used in this work are directly estimated from clinical RT3D data instead of using common parameters from 2D transducers. Based on this phantom, simulated RT3D ultrasound data sets are generated for validation to assess the performance of an optical flow based method in tracking myocardial tissues.

Comparing optical-flow based methods for quantification of myocardial deformations on RT3D ultrasound

Quantitative analysis of cardiac motion is of great clinical interest in assessing ventricular function. Real-time 3-D (RT3D) ultrasound transducers provide valuable three-dimensional information, from which quantitative measures of cardiac function can be extracted. Such analysis requires segmentation and visual tracking of the left ventricular endocardial border. Previously, we presented a method based on four-dimensional optical flow motion estimation for temporal tracking of ventricular borders in RT3D ultrasound. A myocardial displacement field and dynamic cardiac metrics were computed by interpolating the boundary tracking results. In this paper, we propose three additional methods for deriving dynamic cardiac information from tracking ventricular surfaces and demonstrate these methods on a clinical dataset

Comparison study of clinical 3D MRI brain segmentation evaluation

Although numerous methods to segment brain MRI for extraction of white matter, gray matter and cerebrospinal fluid (CSF) have been proposed for the past two decades, little work has been done to evaluate and compare the performance of different segmentation methods on real clinical data sets, especially for CSF. This study focuses on the comparison of the four following methods for segmentation of cerebral brain MRI: gray levels thresholding, three-dimensional level set, fuzzy connectedness and FSL. Quantitative evaluation of segmentation accuracy was performed with comparison to manual segmentation on a database of 10 adult subjects.

Surface Function Actives

Deformable models have been widely used in image segmentation since the introduction of the snakes. Later the introduction of level set frameworks to solve the energy minimization problem associated with the deformable model overcame some limitations of the parametric active contours with respect to topological changes by embedding surface representations into higher dimensional functions. However, this may also bring in more computational load so that recent advances in spatio-temporal resolutions of 3D/4D imaging raised some challenges for real-time segmentation, especially for interventional imaging. In this context, a novel segmentation framework, Surface Function Actives (SFA), is proposed for real-time segmentation purpose. SFA has great advantages in terms of potential efficiency, based on its dimensionality reduction for the surface representation. Utilizing implicit representations with variational framework also provides flexibility and benefits currently shared by level set frameworks. An application for minimally-invasive intervention is shown to illustrate the potential applications of this framework.

Evaluation of optical flow algorithms for tracking endocardial surfaces on three-dimensional ultrasound data

With relatively high frame rates and the ability to acquire volume data sets with a stationary transducer, 3D ultrasound systems, based on matrix phased array transducers, provide valuable three-dimensional information, from which quantitative measures of cardiac function can be extracted. Such analyses require segmentation and visual tracking of the left ventricular endocardial border. Due to the large size of the volumetric data sets, manual tracing of the endocardial border is tedious and impractical for clinical applications. Therefore the development of automatic methods for tracking three-dimensional endocardial motion is essential. In this study, we evaluate a four-dimensional optical flow motion tracking algorithm to determine its capability to follow the endocardial border in three dimensional ultrasound data through time. The four-dimensional optical flow method was implemented using three-dimensional correlation. We tested the algorithm on an experimental open-chest dog data set and a clinical data set acquired with a Philips iE33 three-dimensional ultrasound machine. Initialized with left ventricular endocardial data points obtained from manual tracing at end-diastole, the algorithm automatically tracked these points frame by frame through the whole cardiac cycle. A finite element surface was fitted through the data points obtained by both optical flow tracking and manual tracing by an experienced observer for quantitative comparison of the results. Parameterization of the finite element surfaces was performed and maps displaying relative differences between the manual and semi-automatic methods were compared. The results showed good consistency between manual tracing and optical flow estimation on 73% of the entire surface with fewer than 10% difference. In addition, the optical flow motion tracking algorithm greatly reduced processing time (about 94% reduction compared to human involvement per cardiac cycle) for analyzing cardiac function in three-dimensional ultrasound data sets.

Simulation of 3D ultrasound with a realistic electro-mechanical model of the heart

This paper presents a first set of experiments to integrate a realistic electro-mechanical model of a beating heart into simulated real-time three-dimensional (RT3D) ultrasound data. A novel ultrasound simulation framework is presented, extended from the model of Meunier [12]. True three-dimensional transducer modeling was performed, using RT3D acquisition design. Myocardium and blood scattering parameters were defined in three dimensions. Ultrasound data sets were generated for a normal case and a pathological case, simulating left bundle branch block. Accuracy of an optical flow tracking method was evaluated on the simulated data to measure displacements on the myocardial surfaces and inside the myocardium over a cardiac cycle. The proposed simulation framework has important motivations in a cardiac modeling context as part of this project is focused on the design of effective parameter estimation methods, based on cardiac imaging.

Segmentation of fetal 3D ultrasound based on statistical prior and deformable model

A statistical variational framework is proposed for the fetus and uterus segmentation in ultrasound images. The Rayleigh and exponential distributions are used to model the pixel intensity. An energy is derived to perform an optimal partition of the 3D data into two classes corresponding to these two distributions, in a Bayesian MAP framework. Some numerical difficulties are raised by the combination of heterogeneous distributions in a variational level-set formulation, as discussed in the paper. Results on simulated and real data are presented and show that assuming different distributions provides better results than with the sole Rayleigh distribution.