A Framework for Comparison and Evaluation of Nonlinear Intra-Subject Image Registration Algorithms

Performance validation of nonlinear registration algorithms is a difficult problem due to the lack of a suitable ground truth in most applications. However, the ill-posed nature of the nonlinear registration problem and the large space of possible solutions makes the quantitative evaluation of algorithms extremely important. We argue that finding a standardized way of performing evaluation and comparing existing and new algorithms currently is more important than inventing novel methods. While there are already existing evaluation frameworks for nonlinear inter-subject brain registration applications, there is still a lack of protocols for intra-subject studies or soft tissue organs. In this work we present such a framework which is designed in an ”open-source” and ”open-data” manner around the Insight Segmentation & Registration Toolkit. The goal of our work is to provide the research community with the basis framework that should be extended by interested people in a community effort to gain importance for evaluation studies. We demonstrate our proposed framework on a sample evaluation and release its implementation and associated tools to the public domain.