MatchPoint: On Bridging the Innovation Gap between Algorithmic Research and Clinical Use in Image Registration

摘要

Registration is a basic and crucial step in many clinical image processing tasks. Despite its importance, the evaluation of registration algorithms and transfer into clinical practice are still serious problems. In order to develop an integrated application platform for diagnostics and radiotherapy at the German Cancer Research Center this problems have to be tackled. In this context several problems were identified that are not handled by existing registration libraries (e.g. ITK). These problems unnecessarily handicap the evaluation and transfer into clinical routine. A set of design paradigms were derived in order to handle the problems. This article presents these paradigms and introduces their concrete implementation: MatchPoint. It is a conceptual framework programmed in C++. It aims to amend existing registration libraries by offering an easy integration and access to its design features. Ultimately it aims to ease the handling for non experts, to reduce nonconforming use or resource wastage and to foster efficient integration of registration techniques.