Files and links (1)
PDFOpen Access (License Unspecified), Open Access
Dissertation
Geometric curve/surface-based techniques for curve/surface alignment with application in brain mapping
Doctor of Philosophy (Ph.D.), Drexel University
1999
DOI:
https://doi.org/10.17918/00000444
Abstract
Alignment and registration is an important tool for construction of high-resolution 2D/3D atlases, for quantitative analysis of autoradiograms, as well as for fusing functional and anatomic information to test hypotheses about structural-functional-behavioral correlates in populations and individuals. The alignment is complicated by the variability that exists between animal brains, by transformation effects resulting from the way the data is collected (e.g., freezing the brain), by uncertainties (noisy measurement), by the presence of tears, deformations and incomplete data. These led us to seek a curve/surface representation (B-spline) that is locally controllable, preservable under rigid transformations and shear (the affine group), robust to noise, and continuous with continuous higher order derivatives (at least up to second order for accurate curvature computation). These desirable properties help extracting landmarks on the brain surface and the brain sectional contours. These landmarks are intrinsic, local in nature, and self-preserving under the affine group. The rat brain surface has many curvature variations that ensure the availability of a sufficient number of such landmarks. A whole brain would typically have more than 200 inflection points on its umbilical curves, and a sectional contour would have more than 30 inflection points. To establish correspondences between the landmarks on the atlas and the animal brain (2D contour or 3D surface), a set of absolute geometric 2D/3D invariants are derived. Once the correspondences were established, the affine transformation parameters were estimated and the brain sections/surface is aligned to the atlas. The performance of our method is demonstrated by the ability to register both partial and complete intra and inter-brain contours and surfaces. This compares favorably with the alignment done by a neuroscientist who used both the gray scale information of the stained sections as well as the shape of the sections external contours to base his/her decision.
Metrics
13 File views/ downloads
16 Record Views
Details
- Title
- Geometric curve/surface-based techniques for curve/surface alignment with application in brain mapping
- Creators
- Walid S. Ibrahim Ali
- Contributors
- Fernand S. Cohen (Advisor)
- Awarding Institution
- Drexel University
- Degree Awarded
- Doctor of Philosophy (Ph.D.)
- Publisher
- Drexel University; Philadelphia, Pennsylvania
- Number of pages
- x, 124 pages
- Resource Type
- Dissertation
- Language
- English
- Academic Unit
- Electrical and Computer Engineering; College of Engineering; Drexel University
- Other Identifier
- 991014970315404721