AbstractsPhysics

Intraoperative ultrasound strain imaging displays brain tissue stiffness rather than difference in mass density as regular B-mode imaging. A glioma (the most common type of brain tumor) has a tendency of infiltrating surrounding tissue possibly making it difficult to delineate in a B-mode scan. Complementing B-mode imaging with strain imaging may give additional information about the borders of the tumor thus aiding the surgeons in the task of brain tumor segmentation. It has already been shown that the natural pulsation of the arteries in the brain causes sufficient stress for the displacement in tissue to be measurable with ultrasound.Strain can be estimated along all three spatial dimensions, but is usually estimated in the direction of propagation of the ultrasound beam. Estimating strain in only one direction does not fully characterize the three dimensional movement of the brain tissue. An improvement is to also estimate the strain in the lateral direction. The reason the strain in the lateral direction is often ignored is due to the poor spatial resolution and the lack of phase information.Strain is the spatial derivative of a displacement vector or in the case of two-dimensional strain imaging a displacement map generated by a speckle tracking algorithm. The speckle tracking algorithm performs a correlation search to track a two-dimensional kernel through an image series. A correlation coefficient is calculated for every estimate as an indicator of the accuracy of the estimate. A graphical user interface has been developed to allow the user to change the estimation parameters and view the corresponding results as a cineloop. The displacement estimation is the most time-consuming step of the signal processing chain and makes the method unfit for real-time viewing. Once the displacement is estimated, it is possible to change the strain estimation parameters and see those results in real time thus minimizing the time needed for strain estimation parameter optimization.The speckle tracking method was tested on simulated ultrasound images with a known displacement and on an elasticity phantom. The developed algorithm estimates the displacement and strain in the axial direction with a high accuracy while the estimates in the lateral direction are noisier. The lateral estimates were expected to be less accurate but the results indicate that there might be an error in the simulated images when the scatterers were displaced laterally. For the elasticity phantom, the method was only tested on images acquired while the phantom was axially compressed. It is difficult to decide if the lateral strain estimates shown in the images is actual strain or just noise.The conclusion is that the developed method is able to differentiate structures of different stiffness, but performs better in the axial direction than in the lateral direction, as expected. A better test, and more testing is needed to decide the final performance of the method.