A Comparison of Phase and Speckle Tracking Registration Methods for Motion Correction during HIFU Treatment

Abstract

The accuracy of treatment monitoring and planning is instrumental to the acceptance of HIFU surgery. The ability to locate, analyze and track a feature of interest during treatment will be affected by patient motion. Additionally, statistical analysis and temperature monitoring algorithms would benefit from the registration of successive frames. In this work two registration algorithms, which have had extensive trials in other imaging applications, are investigated. Their ability to reduce patient respiratory and cardiac motion is within ultrasound sequences, taken during HIFU treatments, is compared. The first algorithm is based on an intensity block matching approach with a similarity measure that incorporates speckle statistics explicitly. The second method registers phase representations of the image with a more general similarity measure. These methods would be expected to succeed on different aspects of the image: phase measurements give weight to features and are rotation and contrast invariant, whereas methods to track speckle are successful in images that lack strong features. In general phase based methods of registration are more robust and have the potential to be extended to multimodality registration (such as MRI to Ultrasound), however in this case tracking speckle may produce better results due to the low signal to noise ratio in ultrasound images taken during HIFU treatments. Numerous examples from HIFU surgery are presented to highlight the performance of each method under a various motion constraints. It is shown that the phase based algorithm is generally superior, except in the close proximity to the skin.