Image registration is relevant for many medical procedures. For CT-guided ablation procedures, integrating the lesion location from diagnostic contrast-enhanced CT (CECT) images in interventional CT images may provide better guidance for the interventional radiologist. The main requirement for such methods is to accurately align the lesion location. This, in general, can not be measured, and often surrogates are used for the assessment. In this work, we present a method that permits the assessment of the accuracy of the lesion localization, i.e. assessing the value that is relevant for clinical practice. To this end, we developed a method that virtually removes the contrast agent from an interventional CT image, use this image for the registration, and use the original CECT image for the assessment. For the experimental evaluation, imaging data of 20 subjects (33 lesions) were used, and the registration accuracy of a publicly available registration method was assessed using this method.

doi.org/10.1007/978-3-319-92258-4_8, hdl.handle.net/1765/108931
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Erasmus MC: University Medical Center Rotterdam

Luu, M.H. (Manh Ha), Boulkhrif, H. (Hassan), Moelker, A, & van Walsum, T.W. (2018). Registration evaluation by de-enhancing CT images. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). doi:10.1007/978-3-319-92258-4_8