Konetschnig, J, Saetzler, Kurt and Elbischger, P (2010) A novel framework for Z-registration of electron microscopy image stacks. In: 34th Workshop of the Austrian Association for Pattern Recognition (AAPR/ÖAGM'10), Zwettl, Waldviertel, Austria. Österreichische Computer Gesellschaft (OCG). 8 pp. [Conference contribution]
Full text not available from this repository.
Recent developments in electron microscopy (EM) indicate that focused ion beam (FIB) in combination with field emission scanning electron microscopy (FESEM) presents promising results in capturing the structure Nanogold R marked histones within mitotic chromosomes. In order to compute a three-dimensional density map of their distribution within the chromosome, the generated image series has to be registered. This work introduces a new registration framework built into OpenCAR, a novel open source software for three dimensional reconstruction of biological structures. The scaleinvariant feature transform (SIFT) is used for identifying landmarks within the images. A modified random sample consensus (RANSAC) algorithm then optimizes the transformation model using the generated landmarks. The registration framework is evaluated using serial block-face scanning electron microscopy (SBFSEM) images of neuronal structures as ground truth for accuracy evaluation and FIB/FESEM images of chromosomes are used for repeatability assessment. Delivering promising results in aligning FIB/FESEM image stacks, the proposed method offers an excellent framework for assisting in the 3D reconstruction process of EM image series.
|Item Type:||Conference contribution (Poster)|
|Faculties and Schools:||Faculty of Life and Health Sciences > School of Biomedical Sciences|
Faculty of Life and Health Sciences
|Research Institutes and Groups:||Biomedical Sciences Research Institute > Molecular Medicine|
Biomedical Sciences Research Institute
|Deposited By:||Dr Kurt Saetzler|
|Deposited On:||04 Oct 2011 09:23|
|Last Modified:||04 Oct 2011 09:23|
Repository Staff Only: item control page