Applications of machine learning methods to neuroimaging problems
Event details
| Date | 21.06.2016 |
| Hour | 10:00 › 11:00 |
| Speaker | Prof. Christos Davatzikos, University of Pennsylvania, Philadelphie, PA (USA) |
| Location | |
| Category | Conferences - Seminars |
Machine learning methods have been increasingly used in the past decade to distill complex imaging data down to few informative indices that allow us to achieve diagnoses and predictions on individuals, rather than performing group analyses. We review our work in this area, starting with past work on the use of support vector machines to derive diagnostic indices of Alzheimer’s Disease, schizophrenia, and brain cancer as well as predictive indices of future cognitive decline and clinical progression. We also review similar work on deriving brain development and aging indices, deviations from which flag abnormal processes.
We continue by addressing the important problem of disease heterogeneity, and discuss current work using probabilistic mapping of imaging data from normal populations to those of diseased populations, aiming to answer the question “In how many and which ways do patients X differ from healthy controls?” rather than “in what way do patients X differ from healthy controls?”.
We showcase these methods in studies of aging, MCI, Alzheimer’s Disease and schizophrenia. Finally, we conclude with unsupervised ways of reducing complex imaging data down to a relatively small and informative set of indices, and showcase these methods on functional connectivity and structural data of brain aging.
Bio: Christos Davatzikos, Ph.D.
Wallace T. Miller Sr., Professor of Radiology,
Secondary appt with Electrical and Systems Engineering,
Director, Center for Biomedical Image Computing and Analytics
http://www.cbica.upenn.edu
Director, Section of Biomedical Image Analysis
http://www.rad.upenn.edu/rad.sbia
Joint Affiliations: Bioengineering and Applied Math graduate groups
University of Pennsylvania
We continue by addressing the important problem of disease heterogeneity, and discuss current work using probabilistic mapping of imaging data from normal populations to those of diseased populations, aiming to answer the question “In how many and which ways do patients X differ from healthy controls?” rather than “in what way do patients X differ from healthy controls?”.
We showcase these methods in studies of aging, MCI, Alzheimer’s Disease and schizophrenia. Finally, we conclude with unsupervised ways of reducing complex imaging data down to a relatively small and informative set of indices, and showcase these methods on functional connectivity and structural data of brain aging.
Bio: Christos Davatzikos, Ph.D.
Wallace T. Miller Sr., Professor of Radiology,
Secondary appt with Electrical and Systems Engineering,
Director, Center for Biomedical Image Computing and Analytics
http://www.cbica.upenn.edu
Director, Section of Biomedical Image Analysis
http://www.rad.upenn.edu/rad.sbia
Joint Affiliations: Bioengineering and Applied Math graduate groups
University of Pennsylvania
Practical information
- Informed public
- Free
Organizer
- Prof. Dimitri Van De Ville