Contact : Alexandre Krupa
Last modification : January, 2009
This work deals with the use of speckle information contained in ultrasound (US) images to control the displacement of a robotized 2D ultrasound probe in such a way to stabilize a moving area of soft tissue within the observed ultrasound B-mode imaging plane. This study has been started in 2006 by Alexandre Krupa during his research sabbatical in the Computer-Integrated Surgical Systems and Technology Engineering Research Center (ERC SISST) at The Johns Hopkins University in collaboration with Prof. Gabor Fichtinger and Prof. Gregory D. Hager.
The method makes only use of the speckle information contained in B-mode 2D US images to estimate both in-plane and out-of-plane rigid motions of a moving soft tissue region target with respect to the observed ultrasound image plane.
In-plane motion (2 translations along the image axes and 1 rotation around the image orthogonal axis) is recovered by an image tracking algorithm and out-of-plane motion is estimating using a novel speckle decorrelation technique. This later consists in estimating, from speckle decorrelation measurements, the signed elevation distance of a grid of patches that are fixed on the US plane containing the target region. Fitting a plane to the 3D locations of the patches allows then to extract the 3 out-of-plane displacement components (2 rotations around the images axes and 1 translation along the elevation direction) of the target plane. Estimation of the in-plane and out-of-plane displacements of the target are then used to compute the relative pose between the target image plane and the observed image plane.
Minimizing this relative pose by the use of a hybrid visual servoing control scheme allows then to automatically perform rigid motion compensation of the moving target region. The hybrid visual servoing consists in compensating the in-plane target motion by an image-based visual servoing technique and the out-of-plane motion by a position-based visual servoing technique.
The main advantage of this method is to not rely on segmentation of structure of interest that is still dificult to perform in US image.