Automatic Detection of Wild-type Mouse Cranial Sutures



AbstractIn the study of craniofacial malformations, the cranial sutures are often of interest. The premature fusion of sutures occurring in e.g. Crouzon and Apert syndrome can lead to asymmetric head shape, enlarged intracranial pressure and blindness. In large population studies of such syndromes, automatic detection of the cranial sutures becomes important. We have previously built a
craniofacial, wild-type mouse atlas from a set of 10 Micro CT scans using a B-spline-based nonrigid registration method by Rueckert et al. Subsequently, all volumes were registered nonrigidly to the atlas. Using these transformations, any annotation on the atlas can automatically be transformed back to all cases. For this study, two rounds of tracing seven of the cranial sutures, were performed on the atlas by one observer. The average of the two rounds was automatically propagated to all the cases. For validation, the observer traced the sutures on each of the mouse volumes as well. The observer outperforms the automatic approach by approximately 0.1 mm. All mice have similar errors while the suture error plots reveal that suture 1 and 2
are cumbersome, both for the observer and the automatic approach. These sutures can be hard to detect with the eye. We still believe that
overall, the errors are not considerable and by qualitatively estimating the accuracy, the automatic sutures are very close to the observer sutures. Our plan is to improve the results by local feature detection methods.
TypeConference paper [Abstract]
ConferenceImage Analysis and In-Vivo Pharmacology
Year2007    Month April
BibTeX data [bibtex]
IMM Group(s)Image Analysis & Computer Graphics