Automated Analysis for Three-Dimensional Stress Echocardiography (LGT.6666)

Research
Cardiovascular diseases are one of the major causes of death in the western world. Improvement of diagnostic techniques is therefore of great clinical importance. Stress Echocardiography (stress echo) is a widely applied, relatively inexpensive, safe and non-invasive diagnostic technique for the detection of myocardial dysfunction and of coronary artery disease. Wall motion of the left heart chamber is visualized with echocardiography (ultrasound imaging) in rest and under different levels of stress. Here, "stress" stands for the elevated or maximal performance of the heart, evoked by physical exercise or by pharmacological means. In the classical, two-dimensional Stress Echocardiography (2DSE), 4 standardized 2D cross sections through the heart are made for 3 stress stages plus a recovery stage. However, 2DSE is hampered by several problems, partly inherent to imperfections in image acquisition; partly because of the limitations and subjectivity of visual interpretation. Therefore, there has been a constant quest for a more quantitative and objective form of stress echocardiography.
Recent developments have shown the possibility of three-dimensional Stress Echocardiography (3D stress echo or 3DSE). 3DSE is based on recently introduced fast 3D ultrasound imaging techniques for the heart. 3DSE has excellent potential since it may overcome most of the inherent problems and limitations of 2DSE. However, we believe an automated analysis is the only way to fulfill the need for a more quantitative, objective form of stress echocardiography based on 3D data. 3D imaging will also offer superior possibilities for such automated analysis, since in many cases 3D imaging can solve the ambiguities often encountered in 2D echo, especially in patients with poor image quality. 3DSE is rapidly gaining research interest now that it becomes technically feasible, but a real clinical breakthrough of the technique will likely occur only if an automated analysis will perfect it into a truly quantitative, objective tool.
This project aims at realizing a generic, fully automated analysis technique for three-dimensional stress echocardiography. We propose to apply and test our analysis methods for two recently developed fast 3D imaging techniques that make 3DSE feasible, each with their distinct advantages: the Fast Rotating Ultrasound (FRU) transducer developed by the Erasmus University Rotterdam in cooperation with Oldelft Ultrasound; and the Philips Medical Systems Live3D (Sonos7500) matrix phased-array system. With suitable preprocessing, our approach is applicable for all types of 3DSE data. 3DSE will benefit greatly from automated analysis. In fact three different tasks are in need of full or partial automation: (1) the automated alignment of the 3D image sequences of different stress stages and of different patients; (2) the automatic detection and tracking of the myocardial outlines in the 3D sequences and quantitative analysis of segmental wall motion; and (3) the classification of wall motion abnormalities of different segments, and the related computer-aided diagnosis (CAD) of myocardial condition and narrowings in the coronary arteries. For these automated analyses we propose an original combination of registration, segmentation and classification techniques. The core element will be a 3D Active Appearance Model (AAM), augmented by registration, wall motion estimation and tracking techniques as well as more localized border detection optimizations. We will realize a prototype system for automated analysis of 3DSE images, capable of quantitative measurements of wall motion and computer aided diagnosis of myocardial function. The methods will be developed in close cooperation with clinical users of both 3D systems and the results will be clinically evaluated extensively. The proposed methods will provide an original solution to the difficult problem of quantification of stress echo. From the technical viewpoint, several important novelties concerning AAM are investigated and combined for multiple purposes (registration, segmentation, classification) within one application.

Utilization
Since there is a large demand and market for stress echo (both in terms of number of studies and cost of equipment) and a constant call for objectivity and quantification, there is a great opportunity for quantitative 3DSE that should be fulfilled as soon as possible. Therefore, an automated analysis approach for 3D stress echo has very important implications. It will likely lead to improved diagnosis at low cost and the fast breakthrough of 3DSE, creating a substantial new market.
Furthermore, the developed analysis approaches may generate spinoffs into related fields such as magnetic resonance imaging (MRI), multi-slice computed tomography (MSCT), 2DSE and others. Direct users of the results will be the manufacturers of 3D echocardiography equipment and those of analytical software. End users will be echocardiology departments in academic and peripheral hospitals, as well as clinical research laboratories.
Although we still have to demonstrate the feasibility of our fundamental concepts, we already have the active support of some industrial users: Philips Medical Systems, TomTec Imaging Systems GmbH, Oldelft Ultrasound B.V., Medis medical imaging B.V. and Heart Core B.V. Their active support is the best guarantee for later dissemination of the results and for timely innovations.