Automatic approximation of image registration parameters

In course of radiation therapy (RT) treatment, many different images are acquired to guide the treatment planning process and the fractionated irradiation sessions. These acquired images need to be brought to a common frame of reference so that the information they contain can be correctly processed. Image registration is a method for finding transformations to define this common frame, and thus it is a central topic in the field of radiation therapy. Automatic image registration substantially speeds up the process, and facilitates complex automated workflows. However, there is a multitude of registration methods, even in our research software toolkit SlicerRT and its base platform 3D Slicer. Furthermore, each of these algorithms have a long list of parameters that need to be correctly set to achieve an acceptable registration in clinically feasible computation time. This is further complicated by the multipe available modalities in use (CT, MR, US, etc.), which all have different acquisition protocols and parameters in each clinic.

We propose creating a software tool that analyses input images for registration and suggests a registration method and optimal registration parameters to achieve an optimal registration result.



Project type: 

Masters project