Klaus Obermayer

3D edge detection to define landmarks for point-based warping in brain imaging

The accurate comparison of inter-individual 3D image datasets of brains requires warping techniques to reduce geometric variations. In this study we use a point-based method of warping with weighted sums of displacement vectors, which is extended by an optimization process. To improve the practicability of 3D warping, we investigate 3D operators as landmark detectors for the applicability to our image datasets. The combined approach was tested on 3D autoradiographs of brains of Mongolian gerbils. The warping function is distance-weighted with landmark-specific weighting factors. These weighting factors are optimized by a computational evolution strategy. Within this optimization process the quality of warping is quantified by the sum of spatial differences of manually predefined registration points (registration error). The described approach combines a highly suitable procedure to detect landmarks in brain images and a point-based warping technique, which optimizes local weighting factors.

3D Reconstruction of Neurons from Confocal Image Stacks and Visualization of Computational Modeling Experiments

In this study we perform precise geometrical 3D reconstructions of high complex neuronal architectures. First, confocal microscopy was used to scan neurons with submicron resolution. Second, we extracted the center-lines and diameters of the neuron by means of our reconstruction method, and third we used these metric data to generate compartment models that were transported into the proprietary format of modeling software such as NEURON or GENESIS. Fourth, routines were developed in the scripting language of the respective modeling program to perform computational modeling, and finally we transferred the modeling results to the visualization program AMIRA (Indeed Visual Concepts GmbH).

Warping with optimized weighting factors of displacement vectors-a new method to reduce inter-individual variations in brain imaging

An accurate comparison of multimodal and/or inter-individual 3D image datasets of brains requires geometric transformation techniques (warping) to reduce geometric variations. Here, a subset of warping techniques, namely point-based warping, is investigated. For this kind of warping landmarks between datasets have to be defined. In large 3D datasets manual setting of landmarks is time-consuming and therefore impracticable. Consequently we approach this problem by investigating fast automatic procedures for determining landmarks, based on Monte Carlo techniques. The combined methods were tested on 3D autoradiographs of the brains of gerbils. The results are evaluated by three different similarity functions. We found that the combined approach is highly applicable in processing brain images.

Measuring the Reliability of Geometries in Magnet Resonance Angiography A Reference for Multimodal Image Registration

Magnet Resonance Angiography (MRA) can be used to register MR images of other types (e.g. functional MRI) acquired in the same imaging session as the angiogram since blood vessels are spatially closely confined features. This is only possible if MRA delivers reliable, reproducible images and does not show major random distortions. Therefore, we examine the reliability of MRA over subsequent scanning sessions using an appropriate distance measure on geometric vasculature models obtained from MR angiograms. Additionally we examine the variance between different specimens in order to value the possibility of interspecimen registration.

Adaptive Threshold Masking

Image bias is a usual phenomenon in MR imaging when using surface coils. It complicates the interpretation as well as the algorithmic postprocessing of such data. We introduce a bias correction algorithm based on homomorphic unsharp masking (HUM) that is applicable on a broad range of image types (as long as fore- and background is separable), simple, fast and requires only minimal user interaction. The results of this new algorithm are superior to HUM, especially with regards to feature separability.

Automatic 3D-graph construction of nerve cells from confocal microscope scans

Revised version published in: nJournal of Electronic Imaging. - 12(1), S. 134-150 (Jan 01, 2003). - doi:10.1117/1.1526102 nTitle: nAutomatic three-dimensional graph construction of nerve cells from confocal microscopy scans