Multimodal histological image registration using locally rigid transforms
2014 (English)In: IEEE Transactions on Biomedical Engineering, ISSN 0018-9294, E-ISSN 1558-2531Article in journal (Other academic) Submitted
Evaluating multimodal histological images is animportant task within cancer diagnosis and research. Newmethods are currently under development, such as multiplexingand destaining/restaining protocols, but comparing data fromconsecutive monomodal sections is still the most common methodfor acquiring multimodal data. To allow for comparison of con-secutive sections, registration of the sections is needed. Becauseof the spatial distance between the sections as well as non-uniform deformations, due to mechanical and chemical stressduring handling and staining, this is not a trivial task. Inthis paper, we confirm that deformable transforms outperformlinear transforms when it comes to registration quality. However,large deformations can result in a poor viewing experience forthe pathologist when evaluating the slides, as local structuresare distorted and may look unnatural. The deformations alsoaffect measurements made on the deformed image. We presenta method for locally approximating the global deformabletransform with a rigid transform, and we introduce a gradeof rigidity term that enables a trade-off between registrationquality and measurement distortion. We use a strategy of dividingthe registration in an offline and online step, which gives usthe possibility to perform the approximation in real-time. Thisability offers the viewer with the possibility to quickly switchbetween a view that has optimal registration and a view wheremeasurements are not distorted and where structures ”lookright”. To facilitate further research within the subject, wepresent a registration tool that provides an intuitive interfacefor making comparisons between global deformable transformsand locally rigid approximations with varying degree of rigidity.
Place, publisher, year, edition, pages
Multimodal registration, digital histopathology, locally rigid transforms
Other Computer and Information Science
Research subject Computerized Image Analysis; Computerized Image Processing
IdentifiersURN: urn:nbn:se:uu:diva-219290OAI: oai:DiVA.org:uu-219290DiVA: diva2:699049