Christian Habeck, PhD
Taub Institute
P & S Box 16
630 West 168th Street
New York, NY 10032
Email: ch629 AT columbia DOT edu
Ongoing Research:
Early Alzheimer’s detection with Arterial Spin Labeling (NIA 5R01AG026114)
Arterial Spin Labeling (ASL) is a recent MRI scanning innovation that can measure cerebral blood flow with absolute quantification without the need for arterial injection of any contrast agents. It is therefore substantially cheaper and more comfortable. We are testing ASL and its suitability as an early systems-level biomarker for Alzheimer’s disease by contrasting 20 healthy elderly control participants and 20 early AD patients. Blood flow patterns that are derived from this comparison will be tested for their diagnostic/prognostic ability in 60 participants with Mild Cognitive Impairment. Follow-up data will be available for further validation.
Multivariate Approaches to Brain Imaging Data Analysis (NIBIB 5R01EB006204)
We apply multivariate techniques both to (1) resting data for the purposes of early detection of Alzheimer's disease and (2) to functional data of healthy volunteers for the purposes of understanding how basic cognitive mechanisms are implemented in the brain.
In contrast to the standard univariate techniques in PET and fMRI data analysis, which do not provide any information about region-by-region correlation in the brain, multivariate techniques identify distributed brain networks underlying cognitive function. With my colleague J. Moeller I am extending the covariance-based Subprofile Scaling Model (SSM), pioneered here at Columbia and successfully applied to PET, to fMRI and EEG data, employing both Monte-Carlo simulations as well as empirical surveys of real-world data sets. This proves useful for a better understanding of the techniques, and ultimately of the brain itself. Questions that are of particular interest which are worth addressing are:
- The choice of design matrix X that is applied prior to any covariance-based decomposition has a large influence on the kind of effects that can be identified; this facet has not been appreciated by the community at large, which continues to use "obvious" choices for X that are inspired by univariate SPM analyses but might be inappropriate for covariance analyses.
- How replicable are the results of different analytic tools? We have large enough data sets at our disposal that can address this issue empirically with split-half analyses, rather than merely relying on estimated p-levels.
- How good is the concordance between different flavors of multivariate analysis tools (ICA, PLS, etc.) and how does it depend on the underlying variance structure of the data analyzed?
Representative Publications
1. C. Habeck, N. L Foster, R. Perneczky, A.r Kurz, P. Alexopoulos, R. A. Koeppe, A. Drzezga, Y. Stern. Multivariate and univariate neuroimaging biomarkers of Alzheimer's disease. NeuroImage 2008; 40: 1503-1515
2. C. Habeck, Y. Stern. Neural network approaches and their reproducibility in the study of verbal working memory and Alzheimer's disease. Clinical Neuroscience Research 2007; 6(6): 381-390
3. J. R. Moeller, C. Habeck, Reciprocal Benefits of Mass-Univariate and Multivariate Modeling in Brain Mapping: Applications to event-related functional MRI, H215O-, and FDG PET. International Journal of Biomedical Imaging 2006; Article ID 79862
4. C. Habeck, J.W. Krakauer, C. Ghez, H. A. Sackeim, D. Eidelberg, Y. Stern, J. R. Moeller. A New Approach to Spatial Covariance Modeling of Functional Brain Imaging Data: Ordinal Trend Analysis. Neural Computation 17 (2005) 1602-45.
Software package for spatial covariance analysis with documentation
http://groups.google.com/group/gcva
Popular science reading
1) HARD SCIENCE OR "TECHNICOLOR PHRENOLOGY"? The Controversy over fMRI
By David Dobbs from Scientific American Mind, April 2005
http://daviddobbs.net/page2/page6/page6.html
This is a nice article about the current challenges of fMRI research, and refers to our work described on this homepage.
2) Also check article in "Symmetry Magazine"