Faculty Bibliography

This paper presents a new methodology for the automatic segmentation and characterization of object changes in time series of three-dimensional data sets. The purpose of the analysis is a detection and characterization of objects based on their dynamic changes. The technique was inspired by procedures developed for the analysis of functional MRI data sets. After precise registration of serial volume data sets to 4-D data, we applied a new time series analysis taking into account the characteristic time function of variable lesions. The images were preprocessed with a correction of image field inhomogeneities and a normalization of the brightness function over the whole time series. This leads to the hypothesis that static regions remain unchanged over time, whereas local changes in tissue characteristics cause typical functions in the voxel's time series. A set of features are derived from the time series and their derivatives, expressing probabilities for membership to the sought structures. These multiple sources of uncertain evidence were combined to a single evidence value using Dempster Shafer's theory. Individual processing of a series of 3-D data sets is therefore replaced by a fully 4-D processing. To explore the sensitivity of time information, active lesions are segmented solely based on time fluctuation, neglecting absolute intensity information. The project is driven by the objective of improving the segmentation and characterization of white matter lesions in serial MR data of multiple sclerosis patients. Pharmaceutical research and patient follow-up requires efficient and robust methods with high degree of automation. Further, an enhanced set of morphometric parameters might give a better insight into the course of the disease and therefore leads to a better understanding of the disease mechanism and of drug effects. The new method has been applied to two time series from different patient studies, covering time resolutions of 12 and 24 data sets over a period of roughly one year. The results demonstrate that time evolution is a highly sensitive feature to detect fluctuating structures.

For the measure of brain activation in functional MRI many methods compute a heuristically chosen metric. The statistic of the underlying metric which is implicitly derived from the original assumption about the noise in the data, provides only an indirect way to the statistical inference of brain activation. An alternative procedure is proposed by presenting a binary hypothesis-testing approach. This approach treats the problem of detecting brain activation by directly deriving a test statistic based on the probabilistic model of the noise in the data. Thereby, deterministic and parameterized models for the hemodynamic response can be considered. Results show that time series models can be detected even if they are characterized by unknown parameters, associated with the unclear nature of the mechanisms that mediate between neuronal stimulation and hemodynamic brain response. The likelihood ratio tests proposed in this paper are very efficient and robust in making a statistical inference about detected regions of brain activation. To validate the applicability of the approach a simulation environment for functional MRI is used. This environment also serves as a testbed for comparative study and systematic tests.

The lack of robust and reproducible methods for object segmentation still impedes the introduction of image postprocessing as widely used routine tools in clinical environments. In this paper, we present new tools for the segmentation of two-and three-dimensional objects from multidimensional image data. Our strategy is twofold: After creating an extended graph description of contour fragments and a tessellation of the image plane which is a fully automatic process running in the background, a user can choose between an interactive and a model-based segmentation procedure. A contour grouping algorithm based on path optimization can be used when full user interaction is required. Interactivity is limited to a few simple and quick operations. Another, region-based method uses a split-and-merge strategy and discrete optimization with global shape criteria. Grouping of primitive region patches is invoked by a contour model and a comparison of shape features. In combination, the two procedures form an efficient slice-propagation technique for the segmentation of volumetric objects from three-dimensional image data.

This paper describes a new model-based segmentation technique combining desirable properties of physical models (snakes), shape representation by Fourier parametrization, and modelling of natural shape variability. Flexible parametric shape models are represented by a parameter vector describing the mean contour and by a set of eigenmodes of the parameters characterizing the shape variation. Usually the segmentation process is divided into an initial placement of the mean model and an elastic deformation restricted to the model variability. This, however leads to a separation of biological variation due to a global similarity transform from small-scale shape changes originating from elastic deformations of the normalized model contours only. The performance can be considerably improved by building shape models normalized with respect to a small set of stable landmarks (AC-PC in our application) and by explaining the remaining variability among a series of images with the model flexibility. This way the image interpretation is solved by a new coarse-to-fine segmentation procedure based on the set of deformation eigenmodes, making a separate initialization step unnecessary. Although straightforward, the extension to 3-D is severely impeded by difficulties arising during the generation of a proper surface parametrization for arbitrary objects with spherical topology. We apply a newly developed surface parametrization which achieves a uniform mapping between object surface and parameter space. The 3-D procedure is demonstrated by segmenting deep structures of the human brain from MR volume data.

A novel bias correction technique is proposed based on the estimation of the parameters of a polynomial bias field directly from image data. The procedure overcomes difficulties known from homomorphic filtering or from techniques assuming an initial presegmented image. The only parameters are a set of expected class means and the standard deviation. Applications to various MR images illustrate the performance.

In this paper we present a method for the fully automatic segmentation of cell nuclei from 3D confocal laser microscopy images. The method is based on the combination of previously proposed techniques which have been refined for the requirements of this task. A 3D extension of a wave propagation technique applied to gradient magnitude images allows us a precise initialization of elastically deformable Fourier models and therefore a fully automatic image analysis. The shape parameters are transformed into invariant descriptors and provide the basis of a statistical analysis of cell nucleus shapes. This analysis will be carried out in order to determine average intersection lengths between cell nuclei and single particle tracks of ionizing radiation. This allows a quantification of absorbed energy on living cells leading to a better understanding of the biological significance of exposure to radiation in low doses.