Automatic contour definition on left ventriculograms by image evidence and a multiple template-based model

On the use of motion-based frame rejection in temporal averaging denoising for segmentation of echocardiographic image sequences

We have recently introduced an algorithm for semi-automatic segmentation of the left ventricular wall in short-axis echocardiographic images (EMBC 30:218-221). In its preprocessing stage, the algorithm uses temporal averaging for image denoising. Motion estimation is used to detect and reject frames that do not correlate well with the set of images being averaged. However, the process of estimating motion vectors is computationally intense, which increases the algorithm's computation time. In this work, we evaluate the viability of replacing the motion estimation stage with less computationally intense approaches. Two alternative techniques are evaluated. The ventricular contours obtained from each of the three algorithm variants were quantitatively and qualitatively compared with contours manually-segmented by a specialist. We show that it is possible to reduce the algorithm's computational load without significantly reducing the segmentation quality. The proposed algorithms are also compared with three other techniques from the literature.

Improved segmentation of echocardiographic images using fusion of images from different cardiac cycles

In this work, an algorithm for the detection of the left ventricular border in two-dimensional long axis echocardiographic images is presented. In its preprocessing stage, images fusion was applied to a sequence of images composed of three cardiac cycles. This method exploits the similarity of corresponding frames from different cycles and produces contrast enhancement in the left ventricular boundary. This result improves the performance of the segmentation stage which is based on watershed transformation. The obtained left ventricle border is quantitatively and qualitatively compared with contours manually segmented by a cardiologist, and with results obtained using seven different techniques from the literature.

Semi-automatic algorithm for construction of the left ventricular area variation curve over a complete cardiac cycle

Background

Two-dimensional echocardiography (2D-echo) allows the evaluation of cardiac structures and their movements. A wide range of clinical diagnoses are based on the performance of the left ventricle. The evaluation of myocardial function is typically performed by manual segmentation of the ventricular cavity in a series of dynamic images. This process is laborious and operator dependent. The automatic segmentation of the left ventricle in 4-chamber long-axis images during diastole is troublesome, because of the opening of the mitral valve.

Methods

This work presents a method for segmentation of the left ventricle in dynamic 2D-echo 4-chamber long-axis images over the complete cardiac cycle. The proposed algorithm is based on classic image processing techniques, including time-averaging and wavelet-based denoising, edge enhancement filtering, morphological operations, homotopy modification, and watershed segmentation. The proposed method is semi-automatic, requiring a single user intervention for identification of the position of the mitral valve in the first temporal frame of the video sequence. Image segmentation is performed on a set of dynamic 2D-echo images collected from an examination covering two consecutive cardiac cycles.

Results

The proposed method is demonstrated and evaluated on twelve healthy volunteers. The results are quantitatively evaluated using four different metrics, in a comparison with contours manually segmented by a specialist, and with four alternative methods from the literature. The method's intra- and inter-operator variabilities are also evaluated.

Conclusions

The proposed method allows the automatic construction of the area variation curve of the left ventricle corresponding to a complete cardiac cycle. This may potentially be used for the identification of several clinical parameters, including the area variation fraction. This parameter could potentially be used for evaluating the global systolic function of the left ventricle.

Left ventricle segmentation in echocardiography using a radial-search-based image processing algorithm

A new left ventricle segmentation method in two-dimensional echocardiography images is proposed. Image processing techniques combined with radial search and temporal information are used to extract the left ventricle boundary. Borders from sequential images are extracted using the proposed method, and a curve illustrating the area variation within a cardiac cycle is presented. Performance evaluation is performed by comparing the borders obtained from the presented method to those manually prescribed by a medical specialist. The new sequential radial search algorithm improved the border extraction from long-axis ultrasound images, specially the ones where the mitral valve was open. Segmentation errors due to low contrast were corrected.

Semi-automatic detection of the left ventricular border

Two semi-automatic methods for the detection of the left ventricular border in two-dimensional short axis echocardiographic images are presented and compared. In these methods, the left ventricular area variation curve is calculated during a complete cardiac cycle after the segmentation of several frames. This allows the evaluation of the cardiovascular dynamics and the identification of important clinical parameters. The algorithms are proposed as several independent modules. The results are validated through the comparison between the semi-automatic continuous boundaries and manuals boundaries sketched by a medical specialist.

Artificial neural network: border detection in echocardiography

Being non-invasive and low cost, the echocardiography has become a diagnostic technique largely applied for the determination of the left ventricle systolic and diastolic volumes, which are used indirectly to calculate the left ventricle ejection volume, the cardiac cavities muscular contraction, the regional ejection fraction, the myocardial thickness, and the ventricular mass, etc. However, the image is very noisy, which renders the delineation of the borders of the left ventricle very difficult. While there are many techniques image segmentation, this work chooses the artificial neural network (ANN) since it is not very sensitive to noise. In order to reduce the processing time, the operator selects the region of interest where the neural network will identify the borders. Neighborhood and gradient search techniques are then employed to link the points and the left ventricle contour is traced. The present method has been efficient in detecting the left ventricle borders echocardiography images compared to those whose borders were delineated by the specialists. For good results, it is important to choose properly the areas to be analyzed and the central points of these areas.

Automated contour detection in X-ray left ventricular angiograms using multiview active appearance models and dynamic programming

This paper describes a new approach to the automated segmentation of X-ray left ventricular (LV) angiograms, based on active appearance models (AAMs) and dynamic programming. A coupling of shape and texture information between the end-diastolic (ED) and end-systolic (ES) frame was achieved by constructing a multiview AAM. Over-constraining of the model was compensated for by employing dynamic programming, integrating both intensity and motion features in the cost function. Two applications are compared: a semi-automatic method with manual model initialization, and a fully automatic algorithm. The first proved to be highly robust and accurate, demonstrating high clinical relevance. Based on experiments involving 70 patient data sets, the algorithm's success rate was 100% for ED and 99% for ES, with average unsigned border positioning errors of 0.68 mm for ED and 1.45 mm for ES. Calculated volumes were accurate and unbiased. The fully automatic algorithm, with intrinsically less user interaction was less robust, but showed a high potential, mostly due to a controlled gradient descent in updating the model parameters. The success rate of the fully automatic method was 91% for ED and 83% for ES, with average unsigned border positioning errors of 0.79 mm for ED and 1.55 mm for ES.

Extraction of left ventricular contours from left ventriculograms by means of a neural edge detector

We propose a method for extracting the left ventricular (LV) contours from left ventriculograms by means of a neural edge detector (NED) in order to extract the contours which accord with those traced by a cardiologist. The NED is a supervised edge detector based on a modified multilayer neural network, and is trained by use of a modified back-propagation algorithm. The NED can acquire the function of a desired edge detector through training with a set of input images and the desired edges obtained from the contours traced by a cardiologist. The proposed contour-extraction method consists of 1) detection of "subjective edges" by use of the NED; 2) extraction of rough contours by use of low-pass filtering and edge enhancement; and 3) a contour-tracing method based on the contour candidates synthesized from the edges detected by the NED and the rough contours. Through experiments, it was shown that the proposed method was able to extract the contours in agreement with those traced by an experienced cardiologist, i.e., we achieved an average contour error of 6.2% for left ventriculograms at end-diastole and an average difference between the ejection fractions obtained from the manually traced contours and those obtained from the computer-extracted contours of 4.1%.

Experimental validation of an automated edge-detection method for a simultaneous determination of the endocardial and epicardial borders in short-axis cardiac MR images: application in normal volunteers

The goal of this study was to put together several techniques of image segmentation to provide a reliable assessment of the left ventricular mass with short-axis cardiac MR images. No initial manual input was required for this process based on region growing, gradient detection, and adaptive thresholding. A comparison between actual mass and automatic assessment was implemented with 9 minipigs that underwent spin-echo MR imaging. Fifteen normal volunteers were studied with a fast-gradient-echo sequence. The automatic segmentation was then controlled by three trained observers. Actual mass and automatic segmentation were strongly correlated (r = .97 with P < .01). For normal volunteers, the standard error of estimation of the automatic assessment (12 g) compared well with the average myocardial mass (120 +/- 30 g) and the interobserver reproducibility of the manual assessment (9 g). These results allow the application of this method to the quantification of the left ventricular function and mass in clinical practice.

Precision of myocardial contour estimation from tagged MR images with a "black-blood" technique

RATIONALE AND OBJECTIVES

The authors determined whether blood presaturation of tagged magnetic resonance (MR) images affects identification of left ventricular endocardial borders.

MATERIALS AND METHODS

Three healthy volunteers underwent MR imaging performed with a breath-hold segmented spoiled gradient-recalled-echo sequence with tissue tagging. Two saturation pulses (in the basal and apical regions) were used to generate black-blood images. Manual segmentation of endocardial contours on black-blood and white-blood images was performed independently by five observers.

RESULTS

Endocardial borders were better identified on black-blood images compared with white-blood images, especially in the early systolic phases. Interobserver variability in contour estimation was significantly higher for white-blood images (P < .001) and was twice that for corresponding black-blood images during early systole. Contour variability appeared to be affected mainly by tag-to-myocardium contrast (P = .009) and myocardium-to-chamber contrast (P = .05).

CONCLUSION

Blood presaturation of tagged MR images improves reliability of contour segmentation.

Edge detection in prostatic ultrasound images using integrated edge maps

OBJECTIVE

We investigated an algorithm to detect grey level transitions with multiple scales of resolution to improve edge detection and localisation in ultrasound images of the prostate.

INTRODUCTION

We had developed a non-analytical operator for prostate contour determination implemented with minimum and maximum filters and locate edges. We implemented a technique for improved determination of boundary parts in prostatic ultrasound images by adjusting the edge detection parameter to signal information.

METHODS

First the influence of prefilter settings and edge detection parameters is investigated in a test image and a real ultrasound image. Then, local standard deviation is used to identify or fewer homogeneous regions that are filtered with course resolution, while areas with larger deviation that grey level transitions occur, which should be preserved using smaller filter sizes to improve edge localisation.

RESULTS

Analysis of images with different filter sizes indicated that areas are merged for increasing filter sizes: less pronounced edges disappear or displace for larger filters. Two scales of resolution lead to an improved localisation of edges when smaller filter sizes are used in areas with an increased local standard deviation.

CONCLUSIONS

This paper illustrates an edge detection method suitable as pre-processing step in interpretation of medical images. By adapting input parameters to signal information, object recognition can be applied in images from different imaging modalities. Also, disadvantages are discussed, resulting in a new application combining a localisation algorithm to find the initial contour and a delineation algorithm to improve the outlining of the resulting contour.

Fusion of short-axis and long-axis cardiac MR images

A method for fusion of the short-axis and long-axis cardiac MR images into an isotropic volume image is introduced. A volume image obtained by this method contains the left ventricular (LV) cavity in one piece, facilitating measurement of its shape and volume. The main goal in this image fusion is to reconstruct the LV cavity in volume form and in high resolution. The accuracy of the method is measured using a synthetic image, and examples of image fusion using real images are presented.

Wall position and thickness estimation from sequences of echocardiographic images

Presents a new method for endocardial (inner) and epicardial (outer) contour estimation from sequences of echocardiographic images. The framework herein introduced is fine-tuned for parasternal short axis views at the papillary muscle level. The underlying model is probabilistic; it captures the relevant features of the image generation physical mechanisms and of the heart morphology. Contour sequences are assumed to be two-dimensional noncausal first-order Markov random processes; each variable has a spatial index and a temporal index. The image pixels are modeled as Rayleigh distributed random variables with means depending on their positions (inside endocardium, between endocardium and pericardium, or outside pericardium). The complete probabilistic model is built under the Bayesian framework. As estimation criterion the maximum a posteriori (MAP) is adopted. To solve the optimization problem, one is led to (joint estimation of contours and distributions' parameters), the authors introduce an algorithm herein named iterative multigrid dynamic programming (IMDP). It is a fully data-driven scheme with no ad-hoc parameters. The method is implemented on an ordinary workstation, leading to computation times compatible with operational use. Experiments with simulated and real images are presented.

Quantitative characterization of electron micrograph image using fractal feature

In this investigation, texture analysis was carried out on electron micrograph images. Fractal dimensions and spatial grey level co-occurrence matrices statistics were estimated on each homogeneous region of interest. The fractal model has the advantages that the fractal dimension correlates to the roughness of the surface and is stable over transformations of scale and linear transforms of intensity. It can be calculated using three different methods. The first method estimates fractal dimension based on the average intensity difference of pixel pairs. In the second method, fractal dimension is determined from the Fourier transformed domain. Finally, fractal dimension can be estimated using reticular cell counting approach. Moreover, automatic image segmentation was performed using fractal dimensions, spatial grey level co-occurrence matrices statistics, and grey level thresholding. Each image was segmented into a number of regions corresponding to distinctly different morphologies: heterochromatin, euchromatin, and background. Fractal dimensions and spatial grey level co-occurrence matrices statistics were found to be able to characterize and segment electron micrograph images.

Left ventricular boundary detection from spatio-temporal volumetric computed tomography images

This paper presents a new automatic technique for left ventricle boundary detection from a set of three-dimensional (3D) computed tomography (CT) volumetric cardiac images. The goals of this paper are to incorporate the temporal information into LV boundary detection, to link the shape modeling and LV boundary detection together, and to provide a compact representation of recovered LV boundaries to cardiac imaging. The proposed technique introduces spatio-temporal boundary detection and iterative model-based boundary refinement to left ventricular boundary extraction. The proposed technique has been applied to two sets of four-dimensional (4D) computed tomography images. Experimental results are compared with the manually edited images.

Methods of graph searching for border detection in image sequences with applications to cardiac magnetic resonance imaging

Automated border detection using graph searching principles has been shown useful for many biomedical imaging applications. Unfortunately, in an often unpredictable subset of images, automated border detection methods may fail. Most current edge detection methods fail to take into account the added information available in a temporal or spatial sequence of images that are commonly available in biomedical image applications. To utilize this information the authors extended their previously reported single frame graph searching method to include data from a sequence. The authors' method transforms the three-dimensional surface definition problem in a sequence of images into a two-dimensional problem so that traditional graph searching algorithms may be used. Additionally, the authors developed a more efficient method of searching the three-dimensional data set using heuristic search techniques which vastly improve execution time by relaxing the optimality criteria. The authors have applied both methods to detect myocardial borders in computer simulated images as well as in short-axis magnetic resonance images of the human heart. Preliminary results show that the new multiple image methods may be more robust in certain circumstances when compared to a single frame method and that the heuristic search techniques may reduce analysis times without compromising robustness.

Contour extraction from cardiac MRI studies using snakes

The author investigated automatic extraction of left ventricular contours from cardiac magnetic resonance imaging (MRI) studies. The contour extraction algorithms were based on active contour models, or snakes. Based on cardiac MR image characteristics, the author suggested algorithms for extracting contours from these large data sets. The author specifically considered contour propagation methods to make the contours reliable enough despite noise, artifacts, and poor temporal resolution. The emphasis was on reliable contour extraction with a minimum of user interaction. Both spin echo and gradient echo studies were considered. The extracted contours were used for determining quantitative measures for the heart and could also be used for obtaining graphically rendered cardiac surfaces.

Tag and contour detection in tagged MR images of the left ventricle

Tracking magnetic resonance tags in myocardial tissue promises to be an effective tool for the assessment of myocardial motion. The authors describe a hierarchy of image processing steps which rapidly detects both the contours of the myocardial boundaries of the left ventricle and the tags within the myocardium. The method works on both short axis and long axis images containing radial and parallel tag patterns, respectively. Left ventricular boundaries are detected by first removing the tags using morphological closing and then selecting candidate edge points. The best inner and outer boundaries are found using a dynamic program that minimizes a nonlinear combination of several local cost functions. Tags are tracked by matching a template of their expected profile using a least squares estimate. Since blood pooling, contiguous and adjacent tissue, and motion artifacts sometimes cause detection errors, a graphical user interface was developed to allow user correction of anomalous points. The authors present results on several tagged images of a human. A fully automated run generally finds the endocardial boundary and the tag lines extremely well, requiring very little manual correction. The epicardial boundary sometimes requires more intervention to obtain an acceptable result. These methods are currently being used in the analysis of cardiac strain and as a basis for the analysis of alternate tag geometries.

Automated myocardial edge detection from breath-hold cine-MR images: evaluation of left ventricular volumes and mass

This paper describes an automated edge detection method for the delineation of the endo- and epicardial borders of the left ventricle from magnetic resonance (MR) images. The feasibility of this technique was demonstrated by processing temporal series of cardiac MR images obtained in 12 healthy subjects and acquired from the apex to the base of the heart in multiple anatomic short axis planes with a breath-hold cine-MR acquisition sequence. This procedure allows the entire heart to be imaged in less than 5 min. The automatic program correctly identified the edges in most cases. In poor contrasted images, a fast and user-friendly interactive procedure was used to correct the border delineation. The proposed method for the contour tracing requires a limited degree of control by the user and thus considerably reduces the tedious and long operator time inherent in the usual manual contour tracing tool. The left ventricular volumes were directly measured from these sets of contours by using the Simpson rule, allowing the end-diastolic volumes (EDV), the end-systolic volumes (ESV), the ejection fraction (EF) and the myocardial mass to be determined. The values measured in this study with the dedicated software were similar to the literature values (EDV = 78.3 ml/m2; ESV = 21.1 ml/m2; EF = 73%). Associated with the ultrafast breath-hold cine-MR imaging, the described edge detection method provides an efficient clinical tool for the direct assessment of cardiac function.

Medical image interpretation: a generic approach using deformable templates

We describe a generic approach to image interpretation, based on combining a general method of building flexible template models with genetic algorithm (GA) search. The method can be applied to a given image interpretation problem simply by training a statistical shape model, using a set of examples of the image structure to be located. A local optimization technique has been incorporated into the GA search and shown to improve the speed of convergence and optimality of solution. We present results from three medical applications, demonstrating that the new method offers significant improvements when compared with previously reported approaches to flexible template matching, particularly the ability to deal with different domains of application using a standard method and the possibility of employing complex multipart models. We also describe how the method can be simply extended to track structures in image sequences and segment three dimensional objects in volume images.

Knowledge-based system for boundary detection of four-dimensional cardiac magnetic resonance image sequences

A strategy for a knowledge-based system to detect the interior and exterior boundaries of the left ventricle from time-varying cross-sectional images obtained by ECG-gated magnetic resonance imaging (MRI) is discussed. The system uses both fuzzy set theory and Dempster and Shafer theory to manage the knowledge and to control the flow of system information for more efficient use of limited computational resources and memory space. The key to the approach is that it performs edge detection on images through integration and unification of knowledge and information from edge candidates on all the slices and phases of the acquired cardiac MRI dataset. The analysis system does not base decisions on individual measurements, but on consensus opinions by combining many knowledge sources, some of which may not be completely accurate.

Two approaches to motion analysis of the ultrasound image sequence of carotid atheromatous plaque

It has been observed, with a sequence of ultrasonic images, that a significant number of carotid atheromatous plaques moved under the influence of cardiovascular forces. The aim of this project was to analyse the movement of the heterogeneous and homogeneous plaques using both the discrete approach and the continuous approach. In order to verify that the movement of plaque is not due to the movement of the transducer or the respiration of the patient, global movement estimation was also carried out. Experimentation on an artificial image sequence showed that both approaches can track the target accurately. These approaches were then applied on clinical image sequences and the results showed that carotid plaque was locally moved as a result of the momentum transfer rather than the global movement of the surrounding normal tissue.

Bayesian estimation of ventricular contours in angiographic images

A method for left ventricular contour determination in digital angiographic images is presented. The problem is formulated in a Bayesian framework, adopting as the estimation criterion the maximum a posterior probability (MAP). The true contour is modeled as a one-dimensional noncausal Gauss-Markov random field and the observed image is described as the superposition of an ideal image (deterministic function of the real contour) with white Gaussian noise. The proposed algorithm estimates simultaneously the contour and the model parameters by implementing an adaptive version of the iterated conditional modes algorithm. The convergence of this scheme is proved and its performance evaluated on both synthetic and real angiographic images. The method exhibits robustness against image artifacts and the contours obtained are considered good by expert clinicians. Being completely data-driven and fast, the proposed algorithm is suitable for routine clinical use.

Delineation of ECT images using global constraints and dynamic programming

A model-based delineation algorithm is presented. It is a flexible model fitting algorithm, approaching contour detection as an optimization problem. An objective function is introduced, which depends not only on local contour features, but also on a global shape constraint. The latter is implemented as the similarity to the instance of a parametric shape model. The algorithm optimizes both the contour points and the parameters of the model. As a result, both global and local characteristics of the contour are determined as a compromise between photometric data and prior knowledge. The method was applied to myocardial perfusion SPECT images, to delineate the entire left ventricle (endocardium and epicardium), including possible regions of reduced perfusion. By adapting the balance between the image data and the shape model, images with different characteristics can be processed, including Thallium-201 and MIBI scans.

Epicardial boundary detection using fuzzy reasoning

A fully automated system for detecting the endocardial and epicardial boundaries in a two-dimensional echocardiography by using fuzzy reasoning techniques is proposed. The image is first enhanced by applying the Laplacian-of-Gaussian edge detector. Second, the center of the left ventricle is determined automatically by analyzing the original image. Next, a search process radiated from the estimated center is performed to locate the endocardial boundary by using the zero-crossing points. After this step, the estimation of the range of radius of a possible epicardial boundary is carried out by comparing the high-level knowledge of intensity changes along all directions with the actual image intensity changes. The high-level knowledge of global intensity change in the image is acquired from experts in advance, and is represented in the form of fuzzy linguistic descriptions and relations. Knowledge of local intensity change can therefore be deduced from the knowledge of global intensity change through fuzzy reasoning.

A model-based four-dimensional left ventricular surface detector

The authors have developed a general model-based surface detector for finding the four-dimensional (three spatial dimensions plus time) endocardial and epicardial left ventricular boundaries. The model encoded left ventricular (LV) shape, smoothness, and connectivity into the compatibility coefficients of a relaxation labeling algorithm. This surface detection method was applied to gated single photon emission computed tomography (SPECT) perfusion images, tomographic radionuclide ventriculograms, and cardiac rotation magnetic resonance images. Its accuracy was investigated using actual patient data. Global left ventricular volumes correlated well, with a maximum correlation coefficient of 0.98 for magnetic resonance imaging (MRI) endocardial surfaces and a minimum of 0.88 for SPECT epicardial surfaces. The average absolute errors of edge detection were 6.4, 5.6. and 4.6 mm for tomographic radionuclide ventriculograms, gated perfusion SPECT, and magnetic resonance images, respectively.