Autoregressive modelling for texture similarity analysis as an approach for Ultrasound calibration mismatching assessment

During an interventional procedure, the US probe placed on a patient provides insight at a spatially de ned image plane. The probes position in the world coordinates can be determined using a tracking device placed on it. To use the US probe tracking with a navigation system, the precise spatial transformation between the image scan plane and the tracker placed on the probe has to be established prior to navigation. Ultrasound probe calibration is used to nd this transformation. Current methods to perform US calibration use a phantom with points or certain shapes inside the phantom. These points can be deducted in the US image coordinates and a tracking marker attached to the phantom helps us to nd their position in the world coordinate system. Using this point correspondence we can nd the spatial transformation between the tracker and the US scan plane.
The motivation for this project is to reduce the need for a sophisticated phantom during an interventional procedure, which is time consuming, occupies additional space and it also creates discomfort for the surgeon. We were inspired by the idea of phantomless ultrasound probe calibration and as a rst step towards this goal we want to use a novel similarity approach to assess miscalibration directly from the structures echogenicities in a US intervention. For that the main goal of this project is to propose a new method for computing similarity between US data based on parametrical modelling of the US texture. The main idea is to process a texture as data resulting from a dynamical process that can be modelled using a parametrical approach and whose parameters are used for computing similarity between the data. Data comparison can then be performed from the parametrical representation and not from the data itself as made by general used approach.