target menu
... / ... / ... / 2D/3D imaging / Automated Analysis of 3D Stress Echocardiogra...
 

Automated Analysis of 3D Stress Echocardiography

Advanced automated methods for analyzing cardiac structure and motion in 3D ultrasound images.

Researchers:  Esther Leung, Gerard van Burken, Mike Danilouchkine, Johannes Schaar, Nico de Jong, Ton van der Steen, Hans Bosch

Stress echo
Cardiovascular diseases affect the lives of many people. So methods to determine cardiac function are very important. One such method is stress echocardiography (stress echo). Images of the heart are made using ultrasound, while the patient is in rest in ‘stress’ stages. Here, stress is a state of heightened workload of the heart, brought on e.g. by exercise. By comparing the cardiac motion in these stages, a cardiologist can tell whether the left ventricle is working properly.

 

Project aims
In this research project, automated methods are developed for analyzing three-dimensional (3D) stress echo. With 3D imaging, the true 3D structure and motion of the heart can be assessed. However, 3D images are more difficult to analyze manually. Automated (computerized) methods may allow easier and more quantitative analysis. This should lead to a more accurate diagnosis.

Automated analysis methods
To automate the analysis, we propose using image alignment (registration), contour detection and tracking (segmentation), and classification techniques. Registration is used to align images of the rest and stress stages. This is needed for comparing the cardiac motion in these stages accurately. Segmentation is used to find the borders of the left ventricle. This gives us quantitative diagnostic information, such as the left ventricular volume and the cardiac motion. Classification is used to automatically distinguish between normal and abnormal motion, which indicates cardiovascular disease.

video_url

Video:

Download video fig2-active-appearance

 
Contour detection with active appearance models.

Statistical modeling
Many of our methods make use of patient models. These models describe the typical statistical variations in shape, motion, and image appearance. The models are generated using a database of images, which have been analyzed by experts. As a result, the automated methods can mimic the experts’ decisions, thus giving an intelligent diagnosis.

video_url

Video:

Download video Fig3-image-sequence


Detected contours in image sequence.

Stress echo software
Besides automated methods, we are also developing new software for use in clinical practice. This software, called 3DStressView, allows side-by-side analysis of rest and stress images.

 Afbeelding


3DStressView, specialized  software for side-by-side analysis of rest and stress (click to see an enlarged version).