Automated 2D ultrasound fusion imaging of abdominal aortic aneurysms

Recent studies reveal the benefits of ultrasound strain imaging and elastography of abdominal aortic aneurysms (AAAs). However, feasibility and reproducibility of these techniques is not trivial due to the low imaging depth and low contrast of the US data. Beam-steering overcomes these problems in superficial arteries, but is not applicable for AAAs. Multiangle acquisition could improve both aortic wall segmentation and strain imaging in a similar fashion. In this study, an automated technique for fusion of two-dimensional images, acquired manually at different positions, was developed and applied to ultrasound data of AAAs (n = 5). It was attempted to acquire images at -45, 0 and 45 degrees. Feature points were detected using a scale-space approach and were clustered based on anisotropy of the neighborhood. Next, an ellipsoid was fit through the remaining points. By registering these ellipses, the three different images were compounded. Initial results reveal that the method is able to perform automated registration. The estimated angle between the left and middle images was -25° +/16° and was 37° +/15° between the middle and right position (n = 4). The ellipsoid fit showed more variation in lateral direction. However, additional features should be considered for registration. Results suggest that automated wall thickness assessment might be possible using the extracted feature points.