Tissue characterization

Diffusion MRI is a multi-parametric imaging technique that captures microstructural tissue changes yielding meaningful parameters for tissue analysis. Major limitations of current diffusion MRI techniques go back to the irreproducible relationship between extracted parameters and the underlying real tissue microstructure, which is based on simplifying and often inadequate modelling assumptions. In close cooperation with leading groups around the world, we have developed and validated novel modelling and reconstruction techniques that yield more specific microstructural information in situations of crossing brain fibers and partial volume effects from tissue and free water [1, 5]. Based on these methods, we could demonstrate how partial volume elimination reveals microstructural alterations that precede dementia [1]. These alterations may prove to be an effective and feasible early biomarker of Alzheimer’s disease.

In addition to our work on diffusion-based microstructural tissue characterization, we have conducted research in morphometric tissue parameters. In this context, we have developed an imaging-based anatomical morphometry approach for the quantification of brain atrophy. In another DFG-funded research project “Quantitative analysis of regional lung movement using high temporal resolution MRI” we infer tissue elasticity and ventilation parameters from spatial-temporal MRI imaging.

For the future, we are working towards a multi-modal parametric integration in order to allow for an exhaustive tissue characterization. For this purpose, we are developing novel machine learning techniques for a tissue characterization based on a large database of human primary brain tumors comprising quantitative data from long-term multi-parametric MRI follow-up and biological assays concerning methylation status (Fig. 1).

Fig1
Fig. 1: (a) Automated segmentation of grade IV glioblastoma based on various different MRI contrasts (yellow: edema, red: tumor, blue: necrosis). (b) Depiction of fiber tracts in the vicinity of a grade IV glioblastoma. The volumetric tracking result (yellow) was overlaid on an axial T2-FLAIR image. Red and green arrows indicate the necrotic tumor core and peritumoral hyperintensity, respectively. In the frontal parts, fiber tracts are still depicted, whereas in the dorsal part, tracts seem to be either displaced or destructed by the tumor.


Analysis of Spatial Patterns

We now aim at analyzing the spatial patterns and dynamics of diseases in order to come to a more comprehensive disease analysis. In the context of diffusion-MRI the reconstruction of neuronal pathways in the brain is a key to meaningful spatial analyses. Thus, we have devoted much of our work to the development of proper tractography and validation techniques, which we consider one of the major challenges for the field as a whole (Fig. 1b). The MITK Global Tractography was a very successful method in several successive international competitions at MICCAI 2011-2013.

We developed, validated and applied brain network and white matter skeleton-based analysis techniques that allow the projection and analysis of multiple tissue parameters onto a spatial representation of the brain ([2, 3], Fig. 2).

Fig2
Fig. 2 Schematic overview of brain neuronal network reconstruction. Starting with anatomical imaging (A), the cortex is segmented (B) and parcellated into different functional units, which define the nodes in the network (C). Neural pathways in the brain are identified with diffusion imaging (D) by means of tractography (E). By fusing the anatomical and diffusion data edge weights are determined (F) and a graph representation is derived.


Validation and dissemination

In order to properly disseminate our work, a clear open-source policy was followed. This led to the development of MITK Diffusion, which is publicly available since December 2011 and is a comprehensive resource for diffusion data analysis.

Validation and proper dissemination of results are of tremendous importance in our scientific field of research and turned into their own area of research within the group. In our cooperation with the medical physics department, we were the first to produce realistic fiber phantoms of various complex configurations.

In addition to these hardware phantoms we also developed a comprehensive framework for the generation of realistic synthetic images [4].

Fig3
Fig. 3 Coronal slices of a diffusion MRI brain dataset: (a) The baseline non-diffusion weighted image as it is produced by the scanner. (b) The Generalized Fractional Anisotropy (GFA) calculated on the basis of the analytically reconstructed q-balls. (c) The DTI-based colormap showing the principal diffusion direction with FA modulated intensities. (d) A 2D slice through the tracts generated by the Gibbs tracking algorithm. (e) The track density image. (f) Demonstration of the partial volume analysis module for quantification of the corpus callosum. The examiner draws a circular ROI, which is then automatically clustered in fiber (red), non-fiber (green), and partial volume (transparent). Quantitative measures (e.g. FA values) can then be extracted for the three classes separately. (g) Volume visualization of the FA. (h) Visualization of a network graph using the connectomics module. (i) Tract-based TBSS results visualization of a part of the corpus callosum in an exemplary group study of 15 Alzheimer patients and 15 healthy controls that were enrolled in a DTI imaging study.
 



1.    K. H. Maier-Hein (né Fritzsche), C.-F. Westin, M. E. Shenton, M. W. Weiner, A. Raj, P. Thomann, R. Kikinis, B. Stieltjes, O. Pasternak. Widespread white matter degeneration preceding the onset of dementia. Alzheimer’s & Dementia, 2014 (accepted, in press).

2.    K. H. Maier-Hein (né Fritzsche), R. Brunner, K. Lutz, R. Henze, P. Parzer, N. Feigl, J. Kramer, H.-P. Meinzer, F. Resch, B. Stieltjes. Disorder-specific white matter alterations in adolescent borderline personality disorder. Biol Psychiat, 75(1):81-8, 2014.

3.    J. G. Goch, B. Stieltjes, R. Henze, J. Hering, L. Poustka, H.-P. Meinzer, K. H. Maier-Hein (né Fritzsche). Quantification of changes in language-related brain areas in autism spectrum disorders using large-scale network analysis. Int J Comput Assist Radiol Surg, 2014 (Epub ahead of print).

4.    P. Meher, F. Laun, B. Stieltjes, K. H. Maier-Hein (né Fritzsche). Fiberfox: Facilitating the creation of realistic white matter software phantoms. Magn Reson Med, 2013. (Epub ahead of print)

5.    K. H. Maier-Hein (né Fritzsche), F. B. Laun, H.-P. Meinzer, B. Stieltjes. Opportunities and pitfalls in the quantification of fiber integrity: What can we gain from Q-ball imaging? Neuroimage, 51(1):242-51, 2010.
 

Editor: Email
Latest Revision: 2014-05-16
zum Seitenanfang/up